From cdc7d5ec262d3ce1b3ffa2012800ab159a7fc3d4 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Sun, 26 Jul 2026 16:30:45 -0500 Subject: results: confirm standard-depth SDIL scaling --- results/oral_a_dynamic_scaling_v2/bp_d20_s10.json | 1926 ++++++ results/oral_a_dynamic_scaling_v2/bp_d20_s11.json | 1926 ++++++ results/oral_a_dynamic_scaling_v2/bp_d20_s12.json | 1926 ++++++ results/oral_a_dynamic_scaling_v2/bp_d20_s13.json | 1926 ++++++ results/oral_a_dynamic_scaling_v2/bp_d20_s14.json | 1926 ++++++ results/oral_a_dynamic_scaling_v2/bp_d32_s10.json | 1986 ++++++ results/oral_a_dynamic_scaling_v2/bp_d32_s11.json | 1986 ++++++ results/oral_a_dynamic_scaling_v2/bp_d32_s12.json | 1986 ++++++ results/oral_a_dynamic_scaling_v2/bp_d32_s13.json | 1986 ++++++ results/oral_a_dynamic_scaling_v2/bp_d32_s14.json | 1986 ++++++ results/oral_a_dynamic_scaling_v2/bp_d56_s10.json | 2106 ++++++ results/oral_a_dynamic_scaling_v2/bp_d56_s11.json | 2106 ++++++ results/oral_a_dynamic_scaling_v2/bp_d56_s12.json | 2106 ++++++ results/oral_a_dynamic_scaling_v2/bp_d56_s13.json | 2106 ++++++ results/oral_a_dynamic_scaling_v2/bp_d56_s14.json | 2106 ++++++ .../clean_kp_d32_s10.json | 3390 +++++++++ .../clean_kp_d32_s11.json | 3390 +++++++++ .../clean_kp_d32_s12.json | 3390 +++++++++ .../clean_kp_d32_s13.json | 3390 +++++++++ .../clean_kp_d32_s14.json | 3390 +++++++++ .../clean_kp_d56_s10.json | 3654 ++++++++++ .../clean_kp_d56_s11.json | 3654 ++++++++++ .../clean_kp_d56_s12.json | 3654 ++++++++++ .../clean_kp_d56_s13.json | 3654 ++++++++++ .../clean_kp_d56_s14.json | 3654 ++++++++++ results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json | 1993 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json | 1993 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json | 1993 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json | 1993 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json | 1993 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json | 2089 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json | 2089 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json | 2089 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json | 2089 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json | 2089 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json | 2281 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json | 2281 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json | 2281 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json | 2281 ++++++ results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json | 2281 ++++++ .../oral_a_dynamic_scaling_v2/dynamic_d32_s10.json | 6861 ++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d32_s11.json | 6861 ++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d32_s12.json | 6861 ++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d32_s13.json | 6861 ++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d32_s14.json | 6861 ++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d56_s10.json | 7293 ++++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d56_s11.json | 7293 ++++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d56_s12.json | 7293 ++++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d56_s13.json | 7293 ++++++++++++++++++++ .../oral_a_dynamic_scaling_v2/dynamic_d56_s14.json | 7293 ++++++++++++++++++++ results/oral_a_dynamic_scaling_v2_gate.json | 434 ++ 51 files changed, 168329 insertions(+) create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d20_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d20_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d20_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d20_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d20_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d32_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d32_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d32_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d32_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d32_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d56_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d56_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d56_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d56_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/bp_d56_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json create mode 100644 results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json create mode 100644 results/oral_a_dynamic_scaling_v2_gate.json (limited to 'results') diff --git a/results/oral_a_dynamic_scaling_v2/bp_d20_s10.json b/results/oral_a_dynamic_scaling_v2/bp_d20_s10.json new file mode 100644 index 0000000..be40447 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d20_s10.json @@ -0,0 +1,1926 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d20_s10.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.5931611743164062 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.0868938487625122 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.8971226428031921 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.7778885945701599 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.6786335867118836 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6193286218452454 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5757880435943603 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5440903312683105 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5151293121910095 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.48734404673576354 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4669687108039856 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4494847783088684 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4346876036834717 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.41638119134902957 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.40544724609375 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39685402420043947 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3820656421279907 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3711299934387207 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3656082340335846 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.35637469809532163 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.34379890973091126 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3376724083614349 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3281803913116455 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.325772928314209 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3251122476005554 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31640369718551636 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.3078167387866974 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3063712885379791 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.3043176890659332 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.29318899517059327 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2969893230342865 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.28463393915176394 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.28339950743198394 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.28023815954208375 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.27279416639328 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.27568209482192996 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.27174111887931823 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.26848419264793394 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2705077851486206 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2673165553188324 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.26172702209472654 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.25425094619750976 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.25917373023986817 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.25326495990753173 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2541495544242859 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2518356916046143 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.24633721846580506 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.24365086873054503 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.24470677953243256 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.24789506358146668 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2398070685005188 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.23706979922294616 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.24296264780044555 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.23281645290374756 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.23357934723854065 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.2331781041908264 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.22801256402015685 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.23290970993041993 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.23065821225166322 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.22679906282424928 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.22682038017272949 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.22473457464218138 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.22327382110118865 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.22223310762405396 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.22379935812950133 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.21559041369438173 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.21529934915542603 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.21883329402923585 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.22011523174285888 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.21722387347221375 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.20989989691734315 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.21839928791999816 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.2178214773082733 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.21104170198917388 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.2127850579881668 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.20992795409202575 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.2128899952697754 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.20661681900978088 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.2113364970588684 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.20116662424087525 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.2126935482406616 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.20222373266220092 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.2054754820728302 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.20748530529022216 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.20367337467193602 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.20234411115169526 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.20475129334926606 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.20535121210098267 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.19755666857719423 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.20118453398704528 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.20222160339832307 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.19699565504074096 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.19969011594295502 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1979825215768814 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.19488126012802123 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.19894480196952818 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.19602030634880066 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.19622403920173645 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.19991986083984375 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.19315400151252746 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.11810991772174835 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.09564235707998275 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.08772315320253372 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.08002907380998135 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.07468463415384292 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.07134280677318573 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.06758677285194396 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.06342902672290802 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.06285212803840637 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.05778833916306496 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.05713593818664551 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.05570734820604324 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.05390458149194718 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.05149371256530285 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.04913031956911087 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.04846608467102051 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.046768938567042354 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.04601905714273453 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.04423191998124123 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.04198693593978882 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.040370389528870586 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.040251134046316146 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.0389943587499857 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.039205354404449465 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.03895686731040478 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.037619036943912505 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.037410120245218274 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.03531876205086708 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.035901645948886875 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.034992165200710294 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.034421283962130546 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.03365837794840336 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.031735848093628886 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.03215932679772377 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.030552071770429613 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.031042995034456254 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.031113614351153375 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.029861332706809043 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.031489992849826814 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.030965046857595442 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.028621582498699425 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.028085049097537996 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.027990950353741644 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.02694048818230629 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.026647208641767502 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.025570737427473067 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.027412166399359703 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.025820470059514045 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.027019581270217894 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.026412904968857766 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.023083320103883743 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.021238262484669684 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.021098268574625254 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.019883530657291412 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.02065510878421366 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.019374354729950426 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.01818314094007015 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.019293924129605292 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.018900417600870134 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.018021618705391883 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.019461361792385577 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.017749703515768052 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.018533262635469436 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.017190337671637534 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.016890453641414643 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.017570440534353256 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.016901298728585243 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.01738222026780248 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.01676479687064886 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.017722861864566803 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.017375362040996552 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.017267267870157955 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.016662543128579854 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.016236347911953927 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.01637208472430706 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.01685452579036355 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.016564920703172684 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.01634139787554741 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.017046256511211395 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.017071440948843955 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.015954129823446274 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.01658636713176966 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.01554379013478756 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.015325010071396828 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.015854518936872484 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.015652005437910555 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.015582690757811069 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.015897895269989967 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.015454589307904243 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.015855857462137937 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.015064318771660328 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.016027428925186396 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.01507074480291456 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.016079791139662267 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.016028457076847552 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.015821820734739304 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.014952434335500002 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.01585860956758261 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.015329708568751812 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.015160232656896115 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9139, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.4057889038085937, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2168637952, + "peak_memory_reserved_bytes": 2753560576, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7233767509460449, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3659.870949983597, + "train_wall_s": 3659.1104316711426 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 811020800000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1216531200000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d20_s11.json b/results/oral_a_dynamic_scaling_v2/bp_d20_s11.json new file mode 100644 index 0000000..a4f9d3d --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d20_s11.json @@ -0,0 +1,1926 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d20_s11.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.5890253071975708 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.0681102643966676 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.8947346664428711 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.7738497592544555 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.6827455723190308 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6263094533538819 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5743846928405761 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.545354908027649 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5089111177253723 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.48864533411026 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.45991603262901304 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4445201693725586 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.42800286180496216 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.42017162298202515 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.404514046459198 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3881463805961609 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.37828115423202513 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.37132898860931396 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3632410418224335 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.35251481077194213 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.348331737575531 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3362991683292389 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3301244709300995 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3276013571071625 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.32256933015823364 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3102932949447632 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.30542383867263795 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3090948212528229 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2989043186664581 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2933028840303421 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.29191150889396666 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2854219739532471 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2827479261398315 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2770504412412643 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2740235175895691 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2794273408746719 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2654110499763489 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.27787258545398713 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.26290007787704467 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.26201774267196654 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.26230826616287234 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.25638621970176695 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.25112105862140655 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.25635018618583677 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2500664324235916 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2456773124742508 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.24968147427558898 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2435785795068741 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.24138833020210265 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.24269859988212586 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2418747193145752 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.23975144834518433 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.2369235245704651 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.23111774366855622 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.22708792262077332 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.23293563953399657 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.23114186976909637 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.22791929310798645 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.2324249429988861 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.22661055345535278 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.22082446521759033 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.22716511926174163 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.22301041826725007 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.2172508221578598 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.22076612358093262 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.22163003018379213 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.21729123920440674 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.2123752697944641 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.21622150636672974 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.2143954321193695 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.21372688132286072 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.21410440593719482 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.20973879222869873 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.21167097803592683 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.20795232857227325 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.20475107276916504 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.210544315700531 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.20803603472709656 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.20895568864822386 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.20512487000465393 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.2031609607887268 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.2039771804189682 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.20611678034305572 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.20027091161727906 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.2048240599632263 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.20126052165031433 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.19803103939533234 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.20028031379699707 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.2015662074279785 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1983739560317993 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1968326487827301 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.20168169801712035 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.19480455247879028 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.19671277943134308 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.19210273389816285 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.19285618713378908 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.19492203742027284 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.19434981331825257 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.19507582139015198 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.19142507471323014 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.1190758519411087 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.09520778061866761 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.08196029770851135 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.0780206311917305 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.0701104521036148 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.06796916201114654 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.06422287811994552 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.06132630166053772 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.06057128002166748 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.05790293801426887 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.054991028313338755 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.052029539381265644 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.05037967072367668 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.047320834403038026 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.04915460280418396 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.046505724177360534 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.0453160272693634 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.04368007310628891 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.042369606881141664 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.041392656331062316 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.04019139940261841 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.04025552866876125 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.03674083878844976 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.037137075303792955 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.03686792676448822 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.03609586319923401 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.035135167926847934 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.034851548824310304 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.03274313803613186 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.03143817056179047 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.03185316552758217 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.03032798087477684 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.03103528890967369 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.03066243744879961 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.02859465773880482 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.030074017980098726 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.03026297565340996 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.028658848004341125 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.02843567987024784 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.027601514946222306 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.02671975164949894 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.02621742024600506 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.02652384592682123 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.025169055819511413 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.026105711579322816 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.02511589459955692 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.02500904251396656 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.022639630281925203 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.02537553949996829 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.023891161518096924 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.021664086827039717 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.020598366127610207 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0198515591147542 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.01889366955667734 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.018652743265628813 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.01903608451038599 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.016755025177001955 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.017123083131611348 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.017767101004719735 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.017852005536556245 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.017813909166157247 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.01664680681347847 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.015953939701616765 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.01625092657327652 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.01604759567975998 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.01749425527960062 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.01630456796705723 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.01643814443230629 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0161067535713315 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.016650082113742828 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.016317403566241263 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.015296509456336498 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0163729490506649 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.01615467127740383 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.01552523871436715 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.015830401016771793 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.016695437537506224 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.01552187870979309 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.014730588321387768 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.016039753459095956 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.015830688494741917 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.015698398219645023 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.015214540584683418 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.015529030277729035 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.014786351009756327 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.015306239144206047 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.015075885464549065 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.015085129404067994 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.014986634009182454 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.01571954905003309 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.015959780560433866 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.014082200974822044 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.01487700268343091 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.014305950296074153 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.015252310329079628 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.014214100151360034 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.01365150637805462 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.01587338256418705 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.014280881001651287 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.013817889210954308 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9188, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3808812194824219, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2168637952, + "peak_memory_reserved_bytes": 2753560576, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7187020778656006, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3695.218107700348, + "train_wall_s": 3694.4598400592804 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 811020800000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1216531200000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d20_s12.json b/results/oral_a_dynamic_scaling_v2/bp_d20_s12.json new file mode 100644 index 0000000..97ff7bf --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d20_s12.json @@ -0,0 +1,1926 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d20_s12.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.8098131832504272 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.2565735163116456 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0058431823921203 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8527657214164734 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7465747764778137 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.672649368095398 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6121553597927093 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.574736455745697 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5388540648078919 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.504464215297699 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.485093097820282 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4638041885471344 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.44194923815727233 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.42771271087646484 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4147464679145813 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.40170650484085085 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.39343816712379454 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.37633687685966494 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3739510431480408 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.36397712595939635 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.35377264350891113 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3523500680541992 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3379876541996002 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3280184499549866 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3288205171585083 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31832816835403444 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.3167484021425247 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.30453302896499634 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.30266976757049563 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2997614255619049 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2978201200199127 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.29158487437725067 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.28992508454322813 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2832790825557709 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2777912167072296 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.27536454866409305 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.27246436224937437 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.27243562496185303 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.266952186293602 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2642956964826584 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2574676251888275 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2595961970615387 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2597029899311066 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.25852566350460054 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.25366651141166685 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.24888413954257965 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.24860003258705138 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.25005917612075806 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2410177822494507 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.24534726066589355 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.23348935935020446 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.23628595155715942 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.2373084796333313 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2381118262386322 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.23536265159606934 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.23364131633758545 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.22987226192474366 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.23448671319961548 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.22710167640686035 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.22800665962219238 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.22818440835952758 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.22747452075958252 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.2290425635623932 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.22251493955612184 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.22262729308128357 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.22268718811035157 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.21262741886138917 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.2151704639148712 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.21475640550136565 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.21906636751651765 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.21560630667209627 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.2159178210735321 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.21483301483154296 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.21195224575042726 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.2133840909576416 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.21223120982170104 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.20835534790992738 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.21222200484752654 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.21027848791122436 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.20481170412063598 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.21051985620498656 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.201779447016716 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.2062930567073822 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.20437929859638215 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.20132872997760773 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.20519562193870544 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.20898188190460204 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.2009466459441185 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.19960168691158295 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.20179536536216736 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.20160761443138123 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1970448308300972 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.2001828720808029 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1996461448097229 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.19886296875476836 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.19882571829795836 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.19882662777900695 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.20161866908073425 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1930931982755661 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.19528859308719634 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.12279122568130493 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.09616106805801392 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.08640128634274005 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.07883467857182026 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.07494643726587295 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.07245873218297959 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.06846308543086052 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.0644332435297966 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.06314438662052155 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.060186552579402926 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.056626763789653776 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.055086995103359224 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.05361758773088455 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.051351848690509794 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.049549276272654534 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.04938596292376518 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.04604428948521614 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.04530201344430446 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.0446532622218132 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.045026427310705185 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.040242714388370515 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.04110039208352566 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.041640400770902634 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.038929782221317294 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.037986791943907736 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.03743283396720886 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.03565070045232773 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.03528519847393036 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.03523938384771347 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.035779654843211176 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.03292651256322861 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.031922194584012034 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.03201712161481381 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.031368523983359334 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.030271916363835336 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.031400579726696014 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.030151233365535737 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.030509345423579217 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.030046433401703834 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.027605894335508346 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.02931951781451702 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.029597174162864684 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.027206590920984746 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.026899210656881334 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.026042228419184684 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.026263626831769943 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.027925402495861052 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.026808302974998952 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.026120064418315887 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.025358983798921108 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.023201175934672354 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.021550579267442225 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.021490758321583273 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.01947842873096466 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.01813026362478733 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.019193070428669454 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.018048642708957195 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.01904384822666645 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.01826120739340782 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.018311351767778396 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.018475856048762798 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.01828849165439606 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.018375005535781384 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.01749897241294384 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.01857950935125351 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.018125561828911306 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.017489016358852385 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.01795113804221153 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.017767979103475808 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.01686648118495941 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.017014891899526118 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.01746519201874733 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.016545233925282954 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.01662631719172001 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.016323789744079113 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.016805204996466636 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.017345671919584273 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.01637787770807743 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.015755096770823002 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.01525668834388256 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.017495660351514817 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.01606354157269001 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.016401569648981096 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.015307169351577759 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.015618822374045849 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.016012737891077995 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.015089661456644535 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.014916387542635202 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.01593929616868496 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.014228873479366303 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.01595479544878006 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.01590742797970772 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.015698652393370868 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.015076400068104267 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.016481443066596983 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.015117447131276131 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.014737583790421485 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.015429306066930293 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.015393171034157277 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.014464626797437668 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9142, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3978367706298828, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2168637952, + "peak_memory_reserved_bytes": 2753560576, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7213954925537109, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3660.2713811397552, + "train_wall_s": 3659.514349937439 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 811020800000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1216531200000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d20_s13.json b/results/oral_a_dynamic_scaling_v2/bp_d20_s13.json new file mode 100644 index 0000000..df850be --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d20_s13.json @@ -0,0 +1,1926 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d20_s13.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.631573051223755 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.1040651725578308 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.8767072790908813 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.7480350886535645 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.6605958276176452 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.598413168258667 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5566421480560303 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5198538024711609 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.48770748306274414 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.4618144717216492 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4445196311569214 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.43062618310928347 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.40672400621414184 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.3990583082771301 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.3811617047119141 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3762741340065002 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.36176696568489075 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.35259818826675415 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3451451265144348 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.33508406945228575 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3289540707397461 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3229979333782196 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.31079878032684327 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.30964436631202696 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31385142334461213 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3002555046749115 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2960062663555145 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28911089359760284 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2855690664386749 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2818583485126495 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.278047264623642 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2762251568984985 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.26899996985435487 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.26330243297576905 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2605678303146362 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.26180666986465456 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.25835553429603575 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.25714535642623904 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.25433821109771726 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24957555802345277 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.24520026111602783 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.24539157656669616 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.24716278642654418 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.24432229206323625 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.23490926652431487 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.23579197499275206 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.24048313887596132 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.23130369248390198 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2315685665130615 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.23237737404823303 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.22858958745002747 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2319418381857872 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.22164035197257995 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.22747670996665956 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.22215402113437652 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.22639627418518066 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.21930525390625 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.22228141222000122 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.2215163577747345 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.21842128963470459 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.21542588124275208 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.2127973223876953 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.21363261722564697 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.21342373693943023 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.21423134806632996 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.209122740418911 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.21006568660736083 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.21003352955818178 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.20708554573059082 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.20591702509880067 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.20460648757457733 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.20606652475357057 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.20942618698596954 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.20007646195411682 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.2009643145465851 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.20473031392097474 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.20405819545745849 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.2023987184858322 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.20069762098789215 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.20204433518409728 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.2007487811946869 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1982464499092102 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1947678362083435 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1925799540233612 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.20047719517707824 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.19768478595733643 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.19235636523246766 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.19716073049545288 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.19051602736473083 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.19535972370624544 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1911177098941803 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.19151488564491273 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.19208663902282716 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.19213717898368834 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.19345789896965027 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1838882025051117 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.189594658203125 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.19302816699028016 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.18881871012926102 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1910919116783142 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.11927013695716858 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.09094895141363143 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.08335298265457153 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.07667657580137253 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.07218930123567581 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.06709788583517075 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.06154441981315613 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.05911348286986351 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.056110819444656374 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.055183022159338 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.05337879549741745 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.0488509889292717 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.05093631114006043 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.04915417295217514 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.04489533619999886 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.043104182729721066 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.043134562056064606 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.04339633593440056 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.039702969858646395 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.04072343709468842 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.03727407920002938 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.03841668431758881 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.03651907201886177 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.036206750783920286 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.03434122862696647 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.0342703337675333 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.03399746935129166 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.032454410549402235 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.03202518454194069 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.0319056602987647 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.029810032302737235 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.031136061223745345 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.030901791147589685 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.029925190790891648 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.029190383687019348 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.028014457772374154 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.025982279739379884 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.026057845141291617 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.026730172114372255 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.026647163982391357 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.027202409754395484 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.02624544207930565 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.0247892580640316 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.02314445852935314 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.02476676782026887 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.025546625509262084 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.024034615962207316 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.02481851346373558 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.02465755484640598 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.02333741186261177 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.02010944000005722 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.01940058485344052 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.017991694972217084 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.018560160299241544 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.01813164229899645 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.01711388448834419 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.016111423627734184 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.016526333966255188 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.01658302126854658 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.016772590368688108 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.016276495181024076 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.016098522868156433 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.01597375623732805 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.016056061653196813 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.016393301969468595 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.015538417852818965 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.015548249577814713 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.01552073479294777 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.016039144707918167 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.014914014627337456 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.015560692295432091 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.015365967254042625 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.014784776697456837 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.015537885904312133 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.015172922663092614 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.015521820736676455 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.014627510404288769 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.014950932316482066 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.014838688269257546 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.014139404233694076 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.013770728297531605 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.013919525208473206 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.014313189803361893 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.014399111637920141 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.014686478805840016 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.014232871189117431 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.015125356197357177 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.013718603270053864 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.013918227385282516 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.01378858008235693 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.013887432676255702 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.013475936053991317 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.014294449778795243 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.014533917795419692 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.01412238645374775 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.013795490954220295 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0131240502384305 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.014630534170269965 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.013833322287499904 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.013667404755949974 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9163, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3961239501953125, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2168637952, + "peak_memory_reserved_bytes": 2753560576, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7276139259338379, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3741.015878200531, + "train_wall_s": 3740.255823135376 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 811020800000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1216531200000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d20_s14.json b/results/oral_a_dynamic_scaling_v2/bp_d20_s14.json new file mode 100644 index 0000000..9ee8587 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d20_s14.json @@ -0,0 +1,1926 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d20_s14.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.6420009605026245 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.1364765665054322 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.9320402705574036 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.797639033126831 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.6992854547786712 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6354467618370057 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5860948652458191 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.547577884941101 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5162136293792725 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.4926389773654938 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.47158781294822694 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4482128573226929 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.434623519115448 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4169920855045319 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4057436422252655 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39262936674118043 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.38635027174949643 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3655523226308823 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3588613401889801 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.35510870129585265 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.34988786790847776 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3433113210105896 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3328945175933838 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.323667049074173 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.32010177047729493 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3165242858505249 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.31152601806640623 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3056766118812561 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.300944022436142 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.29420527794837953 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.28958200292587283 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2883981093358994 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2807049588298798 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.28230268796920777 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.27966052257537843 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.27431017803192137 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2686756789445877 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.264711076965332 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2654858276748657 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.25698741995811464 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2600229061603546 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2635122239303589 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2521441179895401 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.2501165414428711 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.24920376007080078 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.24812212656021118 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.24856825354576112 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2494117967700958 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.24060692204475403 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.23912568960666655 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2385230229473114 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.23665064817428588 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.23406595702171326 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2345290622997284 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.23418672790527345 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.23249325196266174 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.2286534098148346 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.22846668304920195 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.22784504460334778 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.22351239710330964 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.2234421351623535 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.2202356809616089 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.22346147294521332 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.22168562659740448 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.2214059203529358 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.21632560173511506 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.220193868560791 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.21736486645698547 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.22082055517196655 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.2095776014328003 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.2123239334487915 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.20440536487579344 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.22123306343078614 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.20609921946525575 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.21385702483177185 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.21147215280532836 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.21294957340240478 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.20560450868606567 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.20830775910377503 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.20710837980747224 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.20211234028339387 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.20526813902378083 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.20310303108215333 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.20534878350257874 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.20346153597831726 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.20476176745414734 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.20210276250362397 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.2029777534198761 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1985911021232605 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.19443689833641054 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.19597420781135558 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.19713825651168823 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.20278231533527374 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.19555891891479493 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.19655887317180634 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.2004914933013916 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.19494393979072572 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1998674341917038 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.19158356195688248 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.19341192204475402 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.12163763278245926 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.09708721893072128 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.08509832964897156 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.07711981454372406 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.07472239377498627 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.06895607421398163 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.06574967886924743 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.06387748464107514 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.06103608752608299 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.05686177908182144 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.05662906150698662 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.05430358715772629 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.051787685266137126 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.05141983794927597 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.04785147648572922 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.04672051605939865 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.04472594115972519 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.04272163626730442 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.04432007931828499 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.042774390975236894 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.04132522941768169 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.03956509531736374 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.03869910090863705 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.03804162994623184 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.03835027703404426 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.03695463723778725 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.0352072454982996 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.035403891465067866 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.03601507543802261 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.03314668522238731 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.03303758738279343 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.032132811360061166 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.030811932396292686 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.03278453142642975 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.03107137560963631 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.029508905804157257 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.029466939091682433 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.029081392669677734 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.029389201905727386 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.029180685257315635 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.027886282885074616 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.027816113124489783 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.02751053996026516 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.026134461442828178 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.027831990154385568 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.026550547634363174 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.026209596066474914 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.024704595115184785 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.025359195474982263 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.024101622509360314 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.021810611231923103 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.020195091626644134 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.018957483512759207 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.018737205794379114 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.019268942840993404 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.017271227710843086 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.017304130091071127 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.018192131876945496 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.01872680107831955 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.017831514892578126 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.01787091297119856 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.016414950402081012 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.017716782972812654 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.017332293102145194 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.01749068653896451 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.01664707453727722 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0164824269887805 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.017310800478756428 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.017300821882486342 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.01572257939636707 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.01685415698543191 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.01567386567234993 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.016720770264863967 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.016096413217782973 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.01740292353928089 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.015035703810453415 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.016655098845362662 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.015033136185407638 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.015558966626524924 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.015409190768301486 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.01679843413054943 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.015670721709430218 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.015452652092576027 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.015030500306487083 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.016188247499465944 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.01531261424511671 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.015122672486305238 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.014619144341647626 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.015147000716924668 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.015481856914758682 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.014748020115494729 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.015293665237873793 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.015706386944949628 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.014918259718120098 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.01502554183959961 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.014863492384329439 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.01511007314413786 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.014687379567325114 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0141090590813756 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.014138477141857148 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.918, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3836814270019531, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2168637952, + "peak_memory_reserved_bytes": 2753560576, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7184293270111084, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3692.4740250110626, + "train_wall_s": 3691.7220034599304 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 811020800000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1216531200000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d32_s10.json b/results/oral_a_dynamic_scaling_v2/bp_d32_s10.json new file mode 100644 index 0000000..37bcaae --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d32_s10.json @@ -0,0 +1,1986 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d32_s10.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.0054121128463747 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4379345008850097 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.1155388502883912 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9032726857757568 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7667027964401245 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.676452765827179 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6164279955673218 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5678001858711242 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5277435734176635 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.4982360699176788 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.47352811789512633 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4530127330207825 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.43370039414405825 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.41464673620223996 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.40234276021003723 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.38951607427597046 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3729274450969696 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.36208275164604187 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.35236940656661986 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3433354270744324 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3330806070423126 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32628141145706174 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3165073899269104 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3097959193992615 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.30136838003158567 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3010846883106232 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2891464224052429 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28535775566101074 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2811079977703094 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.27119437694072723 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27670178437232973 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26804082042694094 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.25923018002510073 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2588696497821808 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2541264269256592 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24870991152763366 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2512641797447205 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2503001854991913 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.24354838258743286 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24228336567401887 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23563844622135163 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2350678129386902 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.23324379225730896 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22955391101837158 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.23024302867889404 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.23181788200378417 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.22251955399513246 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.22411165699958802 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21793724314689636 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21977635389328004 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.21169228483200073 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.21168184700012208 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.21563953169822692 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2104552047586441 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.20921339990615845 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.21239697771072388 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20206186538696289 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.20385388317108155 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.20894759988307954 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.20231014974594116 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.2059004350709915 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.20370832619190216 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.19682449151039125 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.19642020659923554 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.20093582691192627 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.19774235584497452 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18971705436706543 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.19689360201358796 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.19998873517513274 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.19119584038257598 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1885338589167595 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.19239829563617705 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1873906142807007 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.19268658789157866 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1842356293296814 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.18621056265830993 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1835704484939575 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.18719533890724183 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17780928714752198 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.18500971989631654 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.18152961041212082 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.18043139296054841 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.18088321146011352 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.18133888685703278 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.18393014801502228 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.17943879672527313 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1776360273218155 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.17978584024190902 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1760036921596527 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.18229618550300597 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.17563352034807206 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.17081233749866487 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.17763610350608824 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.17541570212841034 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16629342665672303 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.17705314165115357 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.17434777067661286 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16857723032474517 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.17178690128803253 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.17055641087055207 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.1005971980381012 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.07781971687793732 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.06829904849290848 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.06221961904287338 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.05549530576705933 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.05217072459459305 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.04782960817337036 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.046444710136651995 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.04350686324119568 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03984828278422356 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.038114627991318706 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.03725605479836464 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.035026667524576185 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.03393690106987953 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.031377520961761475 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.03009181555867195 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.028700488542318343 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.028414438714981078 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.02713586230635643 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.02669134171426296 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.024967432166934013 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.023414477905631066 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.024176200451254844 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.02272428689658642 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.02198539039373398 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.02231282145857811 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.020591423956751825 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.020472578698694704 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.02021690424859524 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.019907781376838685 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.01926221474170685 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.01843577579677105 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.017993157770186664 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.01773969442784786 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.016086643611192702 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.015963542224168778 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.01700747108578682 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.01534853366971016 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.016440005310773848 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01649655404508114 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.015004711900502444 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.015482185236811639 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.01354793422177434 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.013621482223570347 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.013704320205636323 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.01413012717962265 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.014088259098529815 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.012782217108905316 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.012891866796314716 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.014231026040911674 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.012376270772218704 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.010853191240429879 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.010597322267368436 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.010722288198769093 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.010636799686178564 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.010403271390199661 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.009927612109184265 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.009928871421813965 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.010260844878703356 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.009526269187107682 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.009347029538601637 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.009664224430173635 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.009358572429418564 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.008715892453491688 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.009265132047608494 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.009125266028791665 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.009044252778589725 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.008436540015377103 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.00821531246304512 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.009059850264191628 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.00861273598548025 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.008712154743671418 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.009049842408746482 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.009108233963102102 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0075224792397022245 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.009295342291593551 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.008169813227802514 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.00769507429793477 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.008760777043998242 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.00868443419508636 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.008081732947379352 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.008519896957427263 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.008458584978729486 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.008384161400943994 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.008155007639229297 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.008052484165802598 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.008273763872981071 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.007871800690144301 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.008268070443421602 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.00846611877180636 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.00802742943264544 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.008076108279377221 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.007987720090746879 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.008275099214911461 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.007724166117459535 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.00832881974235177 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.007771533735245466 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.008149257074594497 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.008355577214360237 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.007793944703489542 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9179, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.41951209106445314, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2530507264, + "peak_memory_reserved_bytes": 3254779904, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1608846187591553, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5888.902309417725, + "train_wall_s": 5887.7052166461945 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1377251840000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2065877760000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d32_s11.json b/results/oral_a_dynamic_scaling_v2/bp_d32_s11.json new file mode 100644 index 0000000..102682f --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d32_s11.json @@ -0,0 +1,1986 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d32_s11.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.7063870223617554 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.1592731022262572 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.9175208943557739 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.7536875201416016 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.6544602349090576 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.5927037092971802 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5413562743568421 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5111203736686707 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.4742717270851135 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.45190054489135745 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.42838940790176394 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4102123578834534 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.39549613445281984 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.3842934930706024 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.3674823498725891 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3505688923454285 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3445549236488342 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3351100745010376 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3328804426765442 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3170147942352295 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.31524581239700317 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3041039044237137 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.2955032912158966 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.29121909679412844 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.284248454990387 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.273558822889328 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2708402109146118 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2730033613777161 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.26061281648635864 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.25953510100364685 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.25368027705192564 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.25553169235229495 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.24528157989025115 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2384789893293381 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.23670821279525756 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.23949048082113267 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.23001885991096496 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.23430804176330566 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2266060780906677 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23021400053977967 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2248708538722992 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22127977754592895 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2152890121459961 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.21590086342811585 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21300360562324525 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21362440115451814 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20759890995025634 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2117585801887512 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2054104808998108 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2062998710823059 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.19892980545043945 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2032871570634842 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20141143840789794 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19405519916057587 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1954012571668625 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19321487668991089 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19597827017307282 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.1882704896211624 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19479520617008209 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1866245954322815 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18952527143478393 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18866140976905824 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.19030412703990937 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1900811853647232 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1803061569595337 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18139317154884338 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18498368102550505 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17517292199134826 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1855025179672241 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17939498516082764 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18126282175540925 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.17777664155960082 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17735938474178314 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17590041721343994 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17303487382888794 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17192699808597564 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17615552545547486 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.172336071202755 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1679200311756134 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1712988652896881 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17176300057411195 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.17090996250152587 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16358464079856871 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16778134336471556 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16936078876018523 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16712884270906447 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16496235274791718 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16318997942447663 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1644643171095848 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16694742304325103 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16635796070098877 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15889903512954712 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16751845913887023 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16601602662563325 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1595444133234024 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.15727057287454604 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.15911385097026826 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15316120651245116 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1630015811061859 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1602939554977417 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09227946421146392 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.0663317737698555 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05688851625561714 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.0493025588953495 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04390938655376434 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04130344683885574 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03899301999151707 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03555411759078503 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.0332627472114563 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.032680854793787004 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.031100189917683603 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.029723600989580153 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.027593757073879243 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.026351472999453545 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.023724817582964897 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024201775169968603 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.021525428944826126 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.022219387246370317 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.020941255082786082 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.019570807402133943 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.019837527258992196 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.019317506707310677 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.01744294116437435 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.017821200546994806 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.016505418996214865 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.015808456165790557 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.014928540764003993 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.014331979532986878 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.015485753319859505 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.015043854902088642 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.014684642099738122 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.013334673097729684 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013349158600121737 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.013243361364305019 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.012683310263007879 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.01322338089197874 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.012395091614872218 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.011868331697285175 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012257756915539503 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.011314878867771477 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.011756463255286217 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.010645222712904214 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.010182351255863906 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.011220448665320874 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.009865493751764298 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010065478198081255 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010745289162769914 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.010001082744598388 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.009905803931951522 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.008817061651647091 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008713943128436805 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008190782855153084 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.007377780327275396 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.0077133906853199 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.00847691083431244 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007353016487248242 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007561793578490615 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.006368710192665458 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.006325035137534142 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.006923025486618281 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.00702759523883462 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.0072282628905028104 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.006843369020968676 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006638941957131028 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.006167362957745791 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.006455281130746007 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.006710489913374185 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006123929891437292 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006712332042530179 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006031987841725349 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006468581085950136 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006253111423403025 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006576113955229521 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006163033065497875 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0057193041555956 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006516843093559146 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006436637679915876 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.00553627378128469 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.005849691675044596 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006029486475735903 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006064505338445306 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0064039447094500065 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.005865953974276781 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006197075301744044 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.005685054368823767 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0059968825113028285 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.005682332743108273 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.00587243850775063 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0061469704618304965 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.00573196592360735 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.005621777959764004 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005785522357523441 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005720459768567234 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005167171061113477 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.005886400192156434 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005560370899289847 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.005800981570184231 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.00581084692671895 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005637820402681828 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005325840127561241 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9266, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3595874588012695, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2530507264, + "peak_memory_reserved_bytes": 3254779904, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1661946773529053, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5887.322242021561, + "train_wall_s": 5886.1139216423035 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1377251840000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2065877760000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d32_s12.json b/results/oral_a_dynamic_scaling_v2/bp_d32_s12.json new file mode 100644 index 0000000..8b71383 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d32_s12.json @@ -0,0 +1,1986 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d32_s12.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.0109724760055543 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4362567902374268 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0867284999084472 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8709917178153992 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7273440295028687 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6438600259399414 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5770112455368042 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5361482039642333 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5016758289909363 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.468095518579483 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.44619220394134523 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.42657073191642764 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.40628708760261534 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.38555097502708435 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.379112883605957 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.36289750562667844 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.35174879128456116 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.33422498554229735 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.33106016448020936 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.31727935015678405 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3118976783657074 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3070366426849365 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.29446419066429136 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.28701503961563113 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.2842537051582336 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.27581866315841674 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2763401738643646 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.25907476749420166 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2554603359699249 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.25803864082336425 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2508873863697052 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2454772821187973 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2503407073020935 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.23940209719657898 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.22898770686149597 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.23579959624290467 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.23449487831115723 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.22400421281814575 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.22653502743721007 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.21716231523036958 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22319153874397277 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22315684301137925 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.21558603727817535 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.20974070155858993 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21091635036945344 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2082318727207184 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20161246847629546 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20585539951324464 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2038133184146881 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.19946762784004213 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.19972815490245818 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20028639941215515 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.1937701982116699 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19515884229660035 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19537211338996888 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1926883613014221 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19347582818984985 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19204679753303527 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18423490328788758 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18769514951705932 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18194034811496734 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18289040670871734 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1808374851989746 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18032792008399964 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1827612723827362 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18270389512062074 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18044097856521607 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1740411812543869 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1738271105015278 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1737141721534729 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17230419850826265 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.17472062274932862 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17696084894180297 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16926053122997284 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17055362528800963 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1706799511241913 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.16936428791046143 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.16855762898445129 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16561252969503404 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17293027634620667 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17181514074802398 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16485951325416565 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16485164187431336 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16326760728597642 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.1644478901576996 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.165506713347435 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.164615642824173 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.1611957668232918 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16008895670890808 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16092798224925994 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1590071769285202 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16325321798801423 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.15636748574256898 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16012700378894806 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15904670073986055 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1563712153697014 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.15662149175167084 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15754105580329894 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1567109692144394 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15609369305610657 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08568393441200256 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06317522408127785 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05369775810599327 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.0476024976670742 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.042950796638131145 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.038947348670363424 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.037823320852518084 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.034411116131544114 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03271289059400558 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.0314477956533432 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.029343869721889494 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.028077500077784062 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.026478302937150002 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.026096549355834723 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02390244416564703 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.02377692777991295 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.021567717416882515 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.021640529872179032 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.020908026728034018 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.020051894177794457 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.019267218337357044 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.018688082472383977 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018083301803544165 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.016567399821281432 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.015733772432953118 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.014920587013065815 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.014819765307158232 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.014428511834591626 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.014853895919919013 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.014573111248910427 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.013724009919166565 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.012956059173047542 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.011421179175302386 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.012258051640093326 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.011669609291255474 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.0116468328332901 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.012746921005249023 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.011482126246765256 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.011237891421169042 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.011759749162048102 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.010505393952429295 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011218925432339311 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.01023633032079786 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.010636819964051247 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010898309437930584 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010507725623846054 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010324308806657792 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.009738635150194169 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.008979119207412005 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009272252855747938 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.007726711673140526 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.007999423414170741 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.007616992153301835 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.007475075915306806 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.006915905899107456 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.006768089570701122 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.0068142542649805545 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0067272376950085165 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.006835442341864109 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.0066535683606565 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0068024345889687535 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.006965562088042498 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.006457564204707742 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006979998501986265 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.006785780550315976 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0065035085116326805 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.006269088081642986 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006208638264853507 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0069993398681282995 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006230590722858906 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006384732846021653 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006704631698206067 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006025525711774826 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.005968988159000874 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006251406453922391 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006875649279206991 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006274009262062609 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.005923532918095589 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006215575973615051 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.005989656295031309 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006137610604166985 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.005827424137964845 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.005676852166801691 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.005693613475114107 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.005433482741564512 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006112713556215167 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.005977615090310574 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.005879531293958426 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.005727535793632269 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.005559209338799119 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006220875388532877 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005635415128320455 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006039364632219076 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005313597747534513 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006019642222523689 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005320607864558697 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0062282088583707805 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.005710777210593224 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005980389836505055 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005370809080637992 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9228, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3814170867919922, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2530507264, + "peak_memory_reserved_bytes": 3254779904, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1620919704437256, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5822.728997468948, + "train_wall_s": 5821.533981084824 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1377251840000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2065877760000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d32_s13.json b/results/oral_a_dynamic_scaling_v2/bp_d32_s13.json new file mode 100644 index 0000000..44985ce --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d32_s13.json @@ -0,0 +1,1986 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d32_s13.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.8805035388946534 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.2631674034881593 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 0.982023414478302 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8269909022331238 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7033132957839966 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6245331028175354 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.5728815777492523 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.52921937292099 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.4934953989982605 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.46322856472969054 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4377773000526428 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.42375280329704285 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4023252283382416 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.39270214566230777 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.37170859550476076 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.35657511695861815 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.35467035373687744 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.33651030123710635 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3320449847984314 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.31947522946357726 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.30916595887184145 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.30547217082977296 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.29219140836715696 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.29400309326171875 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.28870507167339327 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2748667608737946 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.27103702613830566 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.26546217414855955 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2625025454044342 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.25851132451057435 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.25504755184173583 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.24713118095397948 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.24437052943229676 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.23974795582294464 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.23825712741851807 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24186695692062377 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.23216649408817291 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.22886585109233856 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.22818136295318603 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2233695182609558 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22007071732521058 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.21992354313850404 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.21873045570373534 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.21697771926403045 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.20397321831226348 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2084566786146164 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20745201335906982 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20686608068466186 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20360097942352295 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2000230358028412 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.19716575583457946 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20097089833259582 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19471816690921784 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.197814011220932 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1954924241065979 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19353147352218628 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.18696056705474853 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19248952531337737 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19309106615543364 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18839726268291473 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1834150011730194 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18410958391189575 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18338790824890136 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1839743905353546 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1801528820514679 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1803589987921715 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1827104704475403 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17917614500522613 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.17485497699737548 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1796244272518158 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1771907129573822 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1721141401386261 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17699346287965775 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17763278216362 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17064625200748443 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.16618630229473114 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17021148312091827 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1735873026251793 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17325835156440736 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17159906162261962 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16759568387508392 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16298270482778549 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16351762491226196 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.166046575756073 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16746359069824218 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16691714178085326 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16248040862560273 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16665220249652862 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16087845814704896 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16221520748138427 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1628304146051407 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15647470529079438 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16293619762420655 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16091344168663024 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15934230563163757 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1541907616329193 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.15469486718654632 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15900839382886886 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16167540033102035 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15679400232315063 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08926333676099778 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06358985314369202 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05420353449106217 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04906956180870533 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.045049789768457416 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.040599410170316694 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03790038890838623 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03717838701725006 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03402421202838421 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03068182231426239 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02941229148209095 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.02811083570063114 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.02811160319328308 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02578263300061226 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.024468044933080673 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.022521375725567342 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.0226251800596714 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.021669687507152556 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.020501725584864617 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.02019407486155629 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.01957855082988739 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01867858926534653 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.017759973801076412 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.017850231526196005 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.016029231131076814 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.015987017872333528 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.016004691554009916 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.015084156606793403 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.015370317788422109 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.014356411390155553 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.013674242456257344 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.01331195755392313 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013287956985011697 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.014606211445629597 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013067925884127618 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.012662553620785475 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.012065714224725962 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.011524995899796485 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.011675733849853277 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01266656897187233 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.010533184029757977 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.010978941943645477 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.010318347289264201 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.010467673452198505 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010728884733617305 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010494428095519542 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010445828742682933 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.009801693223714828 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.010948296121805907 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009674194239675998 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008697000300362707 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.00895926752321422 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008276405196115375 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.007976829496622085 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.0072691182130575184 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007365617476552725 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007068894384503365 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.006627400624677539 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.006712161379158497 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.006901890129819513 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007002437942326069 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007407158368527889 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.006923494810536504 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006925872582197189 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0067213862606883045 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.006998891127705574 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.006689337659217417 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006929537190459668 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006284183492511511 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006663103199824691 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006570300383940339 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006503612499013543 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006347955014780164 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006226098643280565 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0060284089615941045 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006237343201041221 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006058031805083156 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0062861040268838405 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.005802006255611778 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006082883627712726 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006188844832703471 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0057979475711286066 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0063627321705967185 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006196509729810059 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.00635628104865551 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.005718365462571383 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006907654369175434 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.00584992028363049 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.005576714002341032 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.005562419883161783 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006202604257687926 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005533065688759088 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005961296213939786 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005992568579241633 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006175138069316745 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0058324388906359674 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.005814422636181116 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.006527215814217925 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.006064307056963444 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005333981849998236 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9262, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3735280639648437, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2530507264, + "peak_memory_reserved_bytes": 3254779904, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.16585111618042, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5830.384622097015, + "train_wall_s": 5829.188536167145 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1377251840000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2065877760000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d32_s14.json b/results/oral_a_dynamic_scaling_v2/bp_d32_s14.json new file mode 100644 index 0000000..ead9775 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d32_s14.json @@ -0,0 +1,1986 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d32_s14.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.9225823739242554 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4164960892868041 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.120760012626648 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9091575496864319 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7714499942970275 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6778129623222351 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6140075178146363 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5595854544448853 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5196639413070678 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.4925120655441284 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.46359059255599977 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4498627822494507 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4269271025943756 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.40440647114753725 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.38960297353744505 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3763126535797119 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.366042610244751 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3508842104625702 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.34554700902938845 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3303641291618347 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3270281967258453 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.31568570635795595 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3150420280361175 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.30107805528640746 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.2947032307434082 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.294678018283844 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2859807000780106 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2766690892982483 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.27374145025253294 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2638592168903351 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2615679057645798 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.25975301442146304 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2502034167432785 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2520859577322006 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.24862939637184142 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.244263690366745 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24465613820075988 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24179527607917786 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23272236390113832 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22882037929534912 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22537988799095154 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22588210765838623 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22358049662590027 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22093470664978027 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21793900159358978 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21753683715820313 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21588620354175567 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21618832901000976 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21097279445171357 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20663198817253112 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20597183012962342 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.19994518122673036 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20113924821853638 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19977485011100768 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1999194315648079 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19770239804267883 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20051282331466674 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19370327817440033 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19730971799850464 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19116241208076476 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19146727406024933 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.1893305285835266 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18181477528095244 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1908838845062256 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18901293563365937 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1834484317302704 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1857504557991028 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18426053916454316 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1842867259001732 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17477602120399474 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17921741899013519 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.17559958313941956 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1802306512594223 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17615691054821014 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17121027718544007 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17714182065963746 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1763121262741089 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17495869878768922 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17348072474956514 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17383595414400102 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1727399858903885 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1734715978717804 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1673349199104309 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16863740804672242 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16489555994987487 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1702830473613739 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16995898442745208 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.1719565958404541 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16844428910255432 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16289416341781615 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1642383770608902 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16417108777999878 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1624667575740814 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16282382113456725 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16381212590694427 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.15981725409269332 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.15901209310531617 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16209107344150545 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16062442711353303 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.16109401269435883 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09330496985912323 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06679853103995323 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05833959693670273 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.05145497631549835 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04714090671777725 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04397652909398079 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03884652088701725 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03708200654268265 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03475045755326748 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03192536900877953 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.03199194766640663 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.0298711553555727 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.0275455548286438 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.027080157968699933 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.025380560950636864 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.02474314520716667 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.023645476237237454 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02262053836196661 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.021375253983438015 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.02164223212003708 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.021175485536307097 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.020002913225889207 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018995866283774376 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.01745662123799324 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.018009458032250403 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.017360723948776722 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.01653353522092104 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.016291869946420193 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.015165688739567995 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.014390518234670161 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.014876080607920885 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.01354152322858572 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013892504596710206 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.01305307987049222 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013101011054515839 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.011781157428771257 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.013396826519072055 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012641742394864559 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012545095380246639 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01211236124843359 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.012181379710882902 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011682115127891302 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.012197788713127374 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.011751104683354496 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010575061148554086 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010539636700376867 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010397719995826483 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.009707565153241157 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.010903430490493774 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.010372428714781999 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008965822239741683 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008958354243934154 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008743972501903773 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.008012467796206475 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.007939136294722557 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.00772694343701005 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007006239338219165 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.008014692026376724 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.008064465202838182 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.0075252782972157 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.00778471342086792 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.006732449814379215 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007805665935575962 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006874608674496413 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.007625447829514742 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.007315959031283855 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007561641323417425 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.007466862306669354 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.007192313821911812 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.00749102226883173 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006988092204034328 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0068182036101073025 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.005724681941866875 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.007182316865585745 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006710217799684033 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006433702199906111 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006969278431907296 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0060771925245970485 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0069215068834275004 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006596071783900261 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006939981128051877 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006849895872250199 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006535437549203635 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006412262945920229 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006494907861351967 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006900249844565987 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0065344038428366185 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.005925394353866577 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0064781298804283145 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006599408498555422 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006264180855304002 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005788764647431672 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006316005318537355 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005849055158644915 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006100576285738498 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.006544416576102376 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.006031918609589338 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.006268274684920907 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005552465254440904 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005661729123815894 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9216, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3806199676513672, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2530507264, + "peak_memory_reserved_bytes": 3254779904, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1587669849395752, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5810.336062431335, + "train_wall_s": 5809.14767408371 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1377251840000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2065877760000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d56_s10.json b/results/oral_a_dynamic_scaling_v2/bp_d56_s10.json new file mode 100644 index 0000000..b7c6b7e --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d56_s10.json @@ -0,0 +1,2106 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d56_s10.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.194271278152466 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.5154430047988892 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.2268852980804443 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.0365015006256104 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.9015037761306762 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.797037097530365 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7194436560440064 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6643200922012329 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6178795342636109 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5768173001861572 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5424211343574524 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5134369777584076 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.49037210186958313 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4694536312484741 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4519362215042114 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4363633345222473 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.41523621048927306 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.40899277976989745 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.38921398819923403 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.38164856256484986 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3666629805755615 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.35028279891967773 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3400886012840271 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.33685229533195493 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.32847221172332763 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3225873547554016 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.3082633234024048 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3025717590236664 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.29903173513412473 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.294329834690094 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2831515578460693 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2764107035398483 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2736576792621613 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2648253094768524 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2622351321601868 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25753497884750365 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.25615065794944764 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.25152477048397065 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.25019121680259704 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24424193086147308 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23876440581321717 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2342546678543091 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.235261468334198 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.23243886390209198 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22947074512481688 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2292744089984894 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2203837761592865 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21975666457176207 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2119951617336273 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21061336649894713 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2115798826789856 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20938643127918244 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.21179519824028015 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2080950600337982 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.20179885717391968 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.20240866060256957 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20165247373580933 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.1985760055589676 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.2008674271965027 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19240337338447572 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19645147364139556 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18902366543769836 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1956165620803833 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18973047528266906 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1939505928516388 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18665886063575746 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1819526441192627 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18673428000450135 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18326037264823913 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1870834826183319 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18383767890930175 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.18321347157478332 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17582418862342836 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17330053313732147 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17309411842346192 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17790614652633668 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17721375157356262 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17403267610549927 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17467888396263123 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17369678548812867 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17331233839988708 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16721966816425324 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16716315362930298 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.17420471160888673 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16665702794551848 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16273107324123381 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16723198222637176 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16661776196002961 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16578083309173583 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16122940742969513 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16213555507659913 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15995203330993651 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16373665911674498 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16253980697870254 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16031960498809814 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.15767797758102417 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.16427586811065673 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16218511449813844 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1574716680908203 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15869534397125243 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09022786890506744 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06439912135958671 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05144039760351181 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04872830817461014 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.041910421016216275 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.036183351631164554 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03562802261531353 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03186323864102364 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.02950192090690136 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.02901001519858837 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02721487424492836 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.024861279488801957 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.022587017559707163 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02230195553064346 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02122428489148617 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.019177542504668237 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.0181026403722167 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.017672946822047234 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.01618030093342066 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.01577004475623369 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.015246025520563125 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.014032610574066639 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.014495228390991688 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.01440826687425375 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.013247466397285461 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.012899017049223185 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.012293109695017337 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.012777527238726615 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.01209748169824481 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.01114731963276863 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.011745032618641854 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.011201557177901268 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.010488450251817704 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.009851177827566861 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.009776535447686911 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.009974046293944121 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.008849072109162807 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.008943396082594991 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.00936572343841195 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.009284058260768653 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.008175167434066533 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.007842560396045447 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.008325284027550369 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007718517224937677 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.00785332595501095 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.007533050141483546 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.007054277255833149 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006850301388427615 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.007331964514255523 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.007046488393843174 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005963965740054845 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.0062674355255067345 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.00615320375084877 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.00631519119232893 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.006043117364272476 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.0066675276064127684 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.005907805241942406 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.006156193977445364 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.005422134364321828 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.005427361533120275 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.005237732799425721 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.00539790893388912 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.005321698868423701 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.005320854189619422 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.004954823472984135 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.005182363199368119 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.004921885637175292 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.004773076049918308 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.005068077180571854 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.005009522482454776 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.004702887513786554 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.005185032935142517 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.005367844608649612 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.004816462269052863 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.004523273718804121 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.004742496755644679 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.00503995809558779 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0046593975226581095 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0048519244719296695 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.004781357664782554 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.004905174595415592 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.00455825272962451 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.005119736616872251 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.004799448614567518 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0045365674495324495 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.004983578704297543 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.004895393090769648 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.004701349204778671 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.004603644933700562 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.004637074975902215 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.005305963205136359 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.004313608079254627 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.004330640906020999 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.004755494765490293 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.004279542206004262 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.004271870687305927 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.00401694978557527 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.004648360890671611 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.004321102195084095 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.00416296993471682 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9196, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.4136131317138672, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3320109568, + "peak_memory_reserved_bytes": 4542431232, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.075514316558838, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10259.658600568771, + "train_wall_s": 10257.548278093338 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2509713920000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 3764570880000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d56_s11.json b/results/oral_a_dynamic_scaling_v2/bp_d56_s11.json new file mode 100644 index 0000000..e8e6c4a --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d56_s11.json @@ -0,0 +1,2106 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d56_s11.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.2441480759429933 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.6520451573562622 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.2762053594589233 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.0366795889663696 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8838742491340638 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7775179610824585 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7004733442497253 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6399433501815796 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5957588003158569 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5524389961814881 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5249286603546143 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4941679028701782 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4750888855743408 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.45399185165405276 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.43263239362716677 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.41556921830177307 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3966397378349304 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3875173846626282 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3735868306541443 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.358007550201416 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3514554388427734 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3362429839515686 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.328383513584137 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.324324482755661 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3128098294448853 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.30343968832015994 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2968874151802063 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.29214453186035155 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2794385500907898 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2844532305288315 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27004383907318114 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2688684576606751 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2632033786582947 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25779811316013335 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2536117401123047 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24904733840465546 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24340575067520143 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24684725905418395 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23614188342094422 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23577110658168793 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23184648150920867 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22659937084197998 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22489133757591248 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22232432385444642 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21497296612262726 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.213139944896698 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21386680399894714 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21408246718883514 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21116155992031096 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2054360934305191 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20042337512493133 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20099475722312926 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20083012888431548 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.1945902212357521 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19098259191513062 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19665462752342225 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1944172574710846 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.18747816185951233 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19039372524738313 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18545300397872924 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1798163256597519 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18622890206336976 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1808734233045578 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1849400773715973 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.17425392099857331 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.17621845196723937 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.17704346359729767 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17074057980060578 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1723531908416748 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17592787893772124 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.16684336119174958 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1709837682747841 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1664174505519867 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16713098608970642 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1617482682275772 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.16305960576057435 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1654076874256134 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.16277417621135712 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16408164291858673 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15663245958328248 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16177598266601562 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15847330315113067 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15664955557346344 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16187899160861968 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15342998539447786 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.15301464160442352 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15818682832717895 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.15367517629146576 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.15664457752227784 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.15240934684515 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.15182420168876648 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15646429315567018 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.15114465425491333 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14849061069011688 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.14975485221862794 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1467229773426056 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1438916012096405 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15089732598304748 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14869188279151915 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14303171578645707 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07701349678874016 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.052145924034118656 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04184103404998779 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03869001036405564 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.034316612124443056 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.030132720137834548 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.027590993185043336 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.025615437606573106 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.023910948980748654 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.022141397398114205 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.020847919241786004 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01950891440361738 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.01815674560159445 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01754834236353636 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.017942451514899732 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.016193494759500028 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01466585360556841 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.014408022256270052 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.014385275693535805 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.012758618231117725 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.012632293147146702 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.011592723081409932 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.010172025595158339 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.011078532488048076 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.011358492282927036 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.010191038188040256 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.00990184272825718 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.009352294146567583 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.010037282456979156 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009517962433807552 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.009325559734627604 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.00858199987411499 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.009094842192158102 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.009128495811074971 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.008741873390451073 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008326001418977975 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.0077708227922022345 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.008901972435489297 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.0086203098308295 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.007726845970079303 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.0070906995633989575 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.007010017248280346 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.007219308181889355 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007353578768074512 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.007247117817439139 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006711059629842639 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.006540186242870986 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.00603294903755188 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.006158079775236547 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.0055025485432520505 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005286827752552926 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.005041609901227057 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0048378835561685265 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004796428178362548 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004337230419516563 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004345615052171052 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004179104437511415 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.004085922962166369 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0038037374853342773 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.004268488837471232 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004093922774810344 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.004019515627874061 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.004371743677668273 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.004517479064129293 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.004032488136291504 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003401089366339147 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003850532835051417 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.00388622621614486 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.003748435208313167 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0034598920269683004 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0035332952127605675 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003315019896812737 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.003480002392269671 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0035039778316766023 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0036558511624112727 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0035984811946749686 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0035086333359684793 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.003361219122521579 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0036342795138806106 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0031355652825161813 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0031834338897280393 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.003165430543273687 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0033229424182325603 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0033445316010527314 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.003054041137304157 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.003402996248789132 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.003288349488051608 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003205289632026106 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003335792209059 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0031124157635495065 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0034116411933861673 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.002717412325926125 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0032805248819850383 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0030963659314624963 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0032434921885840596 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.003127295686490834 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.00306669891493395 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0028870508106052877 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0027870662285573783 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0027346119079738854 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9275, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.4151858093261719, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3320109568, + "peak_memory_reserved_bytes": 4542431232, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0695407390594482, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10217.597524166107, + "train_wall_s": 10215.493811130524 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2509713920000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 3764570880000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d56_s12.json b/results/oral_a_dynamic_scaling_v2/bp_d56_s12.json new file mode 100644 index 0000000..3593671 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d56_s12.json @@ -0,0 +1,2106 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d56_s12.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.413580471954346 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.8447585605621337 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4722166374969483 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.1879013381576538 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.9690340812492371 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8153593579864502 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7118998365402222 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6407440201950073 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5864878462600708 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5411465914726258 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5059185932540894 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.47923106086730954 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4508411322116852 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.43072596397399904 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4154094973182678 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3904208505821228 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.38186251123428344 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.36342210127830504 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.35131389896392823 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3399477007961273 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.33231564016342163 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32369463687896727 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.31038446986198426 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.30401133107185363 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.2943172943210602 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.28511249803543093 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2835334275484085 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2641226759815216 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.26690744095802305 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2584416003036499 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.25376362020492554 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2517214633560181 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.24121941378593445 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.24271660557746888 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.23234087213516236 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2300577702522278 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.22161671468257904 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.22304309997558594 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.22108167819023133 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.21256760990142823 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2147718205165863 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.20625752338409423 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2088438822555542 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.20263447693824768 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.20056139289855957 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.19413177680969237 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.1955446757864952 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.1940001650285721 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.19066666412591934 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.18382880640029908 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.1866710372209549 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.18865948355674744 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.18220593039512634 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.18327610754013063 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.17775965525627135 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1821687641811371 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1739788373851776 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.17823246103286744 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.17060733947277068 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.16895895728588103 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.17132162360191344 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.1695156877851486 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.17324144656181337 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1637943939781189 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.16513353094100952 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.16646706756591798 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.15926239058971406 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1595205679821968 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.162308915143013 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1565773363161087 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.15622324786663055 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16010785141944886 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.15118239220142365 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.15657217029571532 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.14973657046318053 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.15438437737464905 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1509153999376297 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1547145963525772 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15082019764900206 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1514918416929245 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.14787954893112182 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1492227119064331 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.14918528707504272 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.14154446021556855 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.14837485870361328 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.14523746457576753 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1412196656894684 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14520254972934724 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.13544029650688172 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.14551178688049315 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14038742332458495 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1416707683491707 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14101355106830596 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14156810252428054 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1344923860836029 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14150599776268005 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1339819087791443 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.14047304814338685 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.13870937738895417 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.13526062799453736 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07097873872637749 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.04851436100959778 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.03978079355537891 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03390625799283385 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.029074981991648673 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.026708685238361358 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.025508891676962377 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.022934118223190306 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.02194848285496235 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.01945007309257984 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.018966648765802382 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01751690054371953 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.015639233846515416 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01670257864639163 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.015051313532590866 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.014216155273020267 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.013214815193116665 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.012478619110286237 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.011920008748471737 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.01135440178528428 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.010664004289656878 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01087159504108131 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.010201428257524967 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.009868202875107528 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.008965574307087809 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.00891771702349186 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.008399974834546447 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.007698562731146812 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.008475578688755632 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.00830902346316725 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.007853451461046934 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007851619808003306 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.006879737859144807 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.007379789023473859 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.007331988705322146 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.006300915867090225 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.006078072411343455 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.006132870688810944 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.006521193671487272 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.005947416508868337 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.005325597274452448 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.006016764365881682 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.0060249122547730806 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.005536915111280977 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.004927738820463419 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.005624542062422261 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.005065151788778603 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.005377319272048772 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.005547104981746525 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005009888500422239 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.004688873038478196 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.00445300730606541 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.004359322290997952 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.0036598509731888772 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.003909484916701913 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.003556795653514564 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004289821053119376 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0037766539711505172 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0035305865063518286 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.003906310083977878 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.003510146408043802 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.003772904524244368 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.003453466680981219 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.00404699356328696 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0034331864208355547 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003324918228946626 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0031030424808710815 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.003629748016297817 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0031882802027463912 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.003775105935363099 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.003154012319855392 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003352703639790416 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0031730502455495297 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0028874678939394653 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0032457301169633865 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0034553286668844523 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.003174144511334598 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0033225763526372612 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.003090857371427119 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0030615870848856867 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.003162896712049842 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0029614244273304937 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0027170680877566336 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.002877572488244623 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0029651545545086263 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.003176359420884401 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0031714606587029993 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0025045483745634554 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003140454687066376 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0028794106182456016 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0030680617382749915 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0030037131494283676 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.002765146735534072 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.003065732864001766 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0027103334291651844 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.002674253397770226 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0030751690716296436 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.002720250786393881 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.002744796032812446 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.002791558302119374 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9288, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.368893408203125, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3320109568, + "peak_memory_reserved_bytes": 4542431232, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.064631223678589, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10241.933519124985, + "train_wall_s": 10239.837575912476 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2509713920000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 3764570880000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d56_s13.json b/results/oral_a_dynamic_scaling_v2/bp_d56_s13.json new file mode 100644 index 0000000..ae4be6f --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d56_s13.json @@ -0,0 +1,2106 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d56_s13.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.2380722662353514 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.5243022811508178 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.1791382378959656 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9526055066299438 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7985830499458313 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6951202674484253 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.621785017118454 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5692547700881958 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5245636717605591 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.48866498497009275 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4679339222049713 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.44021042459487914 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4147442305088043 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4020823833656311 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.3830681062889099 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.36699817971229554 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3528458435153961 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3351625449562073 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.33076745034217836 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.314546350107193 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.31316587451934813 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.29657996692657473 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.2838392660522461 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.2806135774803162 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.28083187955856326 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.26689816400527955 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.25746106077194214 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.25681675987720487 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.24887987003326417 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2419551268196106 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.23928709052085875 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.23788902492523192 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.23008561833381652 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.22262225829124452 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.22364863095283508 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.21420050528526305 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2129284870958328 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2079516935443878 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.20886797585010528 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2089265088558197 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.19851134345531463 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.19988643452644347 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.19532940200805665 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.195343414645195 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.19538965277671813 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.1897829840183258 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.19165333584070204 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.18658844998836518 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.18472965229034424 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.1821135854434967 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.17800462348937987 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.18199947471618652 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.1763830384373665 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.1756221105480194 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1696118577528 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1749851853656769 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1657337329673767 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.17245514731407166 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.17474647249221803 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.16800471481323243 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1669558035326004 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.16427249970912933 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1582439007759094 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1585183248090744 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.16085544743061064 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1578859422969818 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.15871727242946626 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.15582639401435852 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.15439380012512208 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.15781796977996826 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.15552019326210023 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1537364743041992 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.157676806306839 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.15155513235092163 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.14866421595573426 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1499950579404831 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.14595741822242736 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1475743013525009 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1440841651725769 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15019753473758699 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.14983026386260986 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1457690325498581 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1425291301584244 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.13966907420635225 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.14920791668653488 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1433650892972946 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.14277835859298707 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14652419395685196 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.13885879600524903 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.13664149361371994 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14250784899711608 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.136539647898674 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.13834525762081146 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.13874304574012755 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.13826945496559143 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.13466810110569 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.13618385932683943 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.13083945737361907 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14239856719732286 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.13760523827075957 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07247622606992722 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.047615873658657076 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.039028446272611615 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03336639288246632 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.030500985317230226 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.02714287847608328 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.023862378755509853 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.02249471425935626 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.019693836632966994 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.018667535536289213 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.017429370289444922 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01614685102917254 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.016091273787617683 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01417996779859066 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.01396479100972414 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.012489625472724438 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.012624856723845005 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.011601389402598142 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.010255140935406089 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.010553088017329574 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.010899415211081506 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.010135487665981054 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.009882493694797158 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.008986017334088683 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.009793174834996463 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.008635873577445746 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.007427890453115106 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.007159797935336828 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.0072894377361983065 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.00702851181268692 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.0071594165313243864 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007001433908045292 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.00688736757075414 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.006680425413325429 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.006584880689047277 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.006494963362403214 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.006134787359386683 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.005662843820005655 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.005598675254136324 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.006541163565441966 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.005306101936548948 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.0051230572059936825 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.005335996469631791 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.005076246513798833 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.00512076975941658 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.005122748474702239 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.004380679072439671 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.004719913615286351 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.004689400273300707 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.004892317248731851 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.00447578532397747 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004113756105890497 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0037811573535203933 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.003723595006912947 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.0036731692541390657 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.003562408784888685 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.00363837756998837 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0036040416952967643 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.003390372982798144 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.0033142458989098666 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.003420861539710313 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.003407911326214671 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.003366744888164103 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.0033962769866734744 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0033505655673518775 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0033576356161758303 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003377419900596142 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.003348771807849407 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0035825366028398274 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.003371795840486884 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.002851168489456177 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.00323815221786499 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0031118369930982588 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0028565359775349496 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.002883928079903126 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.003193774028085172 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0031581616420298816 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0031817424205318095 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0031136344198696314 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0027644987058639526 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.003138273613899946 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0031510034423321487 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.002888658501994796 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0027053265723958613 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.002944691154398024 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.002828227263186127 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.002878027893342078 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.002947899527810514 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0031204521046578884 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0030550059485808015 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0029061227486096323 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.002946553324870765 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.003338872848302126 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.002944037918429822 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.002767612297460437 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0028004104109480977 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0026937156117893754 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0031516437913104893 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0030365199307352303 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0025308626214414834 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9306, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3573267272949219, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3320109568, + "peak_memory_reserved_bytes": 4542431232, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0722508430480957, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10194.431643486023, + "train_wall_s": 10192.328938484192 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2509713920000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 3764570880000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/bp_d56_s14.json b/results/oral_a_dynamic_scaling_v2/bp_d56_s14.json new file mode 100644 index 0000000..f754d60 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/bp_d56_s14.json @@ -0,0 +1,2106 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 0, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "bp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/bp_d56_s14.json", + "output_lr": null, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": null, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.1491518631744384 + }, + { + "epoch": 2, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.5415329328155518 + }, + { + "epoch": 3, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.1902907227325439 + }, + { + "epoch": 4, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9493342241477967 + }, + { + "epoch": 5, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8143947232627868 + }, + { + "epoch": 6, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.733723431892395 + }, + { + "epoch": 7, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6721547922706604 + }, + { + "epoch": 8, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6167750704574585 + }, + { + "epoch": 9, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5720778098678588 + }, + { + "epoch": 10, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5399607049369812 + }, + { + "epoch": 11, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5111437025451661 + }, + { + "epoch": 12, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4847265879821777 + }, + { + "epoch": 13, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4709283294296265 + }, + { + "epoch": 14, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.43931803150177 + }, + { + "epoch": 15, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.42153508222579955 + }, + { + "epoch": 16, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4142340124320984 + }, + { + "epoch": 17, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.40057316608428956 + }, + { + "epoch": 18, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3751462773513794 + }, + { + "epoch": 19, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3673887727355957 + }, + { + "epoch": 20, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.35467099767684934 + }, + { + "epoch": 21, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3524684285068512 + }, + { + "epoch": 22, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.34037952941894534 + }, + { + "epoch": 23, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3340830976486206 + }, + { + "epoch": 24, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3162274563121796 + }, + { + "epoch": 25, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3063911257553101 + }, + { + "epoch": 26, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3132950550365448 + }, + { + "epoch": 27, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29966708919048307 + }, + { + "epoch": 28, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.29376793018817904 + }, + { + "epoch": 29, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28038517884254455 + }, + { + "epoch": 30, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2781383064031601 + }, + { + "epoch": 31, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2694102143526077 + }, + { + "epoch": 32, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2667882744693756 + }, + { + "epoch": 33, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2611733724021912 + }, + { + "epoch": 34, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25603877515792844 + }, + { + "epoch": 35, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25416267645835877 + }, + { + "epoch": 36, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24760894286632537 + }, + { + "epoch": 37, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24410780872821808 + }, + { + "epoch": 38, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24305856939315795 + }, + { + "epoch": 39, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23772317601203918 + }, + { + "epoch": 40, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2337888228416443 + }, + { + "epoch": 41, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22836926259040832 + }, + { + "epoch": 42, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22831247077941894 + }, + { + "epoch": 43, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22156494766235352 + }, + { + "epoch": 44, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.215394432888031 + }, + { + "epoch": 45, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21597371877670288 + }, + { + "epoch": 46, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21680292412757873 + }, + { + "epoch": 47, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2124243455505371 + }, + { + "epoch": 48, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21217001245975495 + }, + { + "epoch": 49, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.207936761264801 + }, + { + "epoch": 50, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20709203666687012 + }, + { + "epoch": 51, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.19351246401786804 + }, + { + "epoch": 52, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20142732095718383 + }, + { + "epoch": 53, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.191461335067749 + }, + { + "epoch": 54, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19068031177997588 + }, + { + "epoch": 55, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1948518556499481 + }, + { + "epoch": 56, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19615788020133973 + }, + { + "epoch": 57, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19002282517433167 + }, + { + "epoch": 58, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.1905314666724205 + }, + { + "epoch": 59, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18444560183048248 + }, + { + "epoch": 60, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.17953283669233322 + }, + { + "epoch": 61, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.17936252091407776 + }, + { + "epoch": 62, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.17894449977874755 + }, + { + "epoch": 63, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18608696917533873 + }, + { + "epoch": 64, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1867162022972107 + }, + { + "epoch": 65, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.17704015726566313 + }, + { + "epoch": 66, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.17832065382003784 + }, + { + "epoch": 67, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1795563365650177 + }, + { + "epoch": 68, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17364684770584107 + }, + { + "epoch": 69, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.17744720470428466 + }, + { + "epoch": 70, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17062861320495606 + }, + { + "epoch": 71, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17449909383296966 + }, + { + "epoch": 72, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1682535918712616 + }, + { + "epoch": 73, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17165554725170135 + }, + { + "epoch": 74, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.15960627580165862 + }, + { + "epoch": 75, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.16947615956306458 + }, + { + "epoch": 76, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.16465355763435363 + }, + { + "epoch": 77, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15931276038169861 + }, + { + "epoch": 78, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.16503633592128752 + }, + { + "epoch": 79, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16429143577575683 + }, + { + "epoch": 80, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.16562639446258545 + }, + { + "epoch": 81, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16291448392152785 + }, + { + "epoch": 82, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15886301523447036 + }, + { + "epoch": 83, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.160158472032547 + }, + { + "epoch": 84, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1616283896446228 + }, + { + "epoch": 85, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16116115434646605 + }, + { + "epoch": 86, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.15407067019462586 + }, + { + "epoch": 87, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15864971215248108 + }, + { + "epoch": 88, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.15416404981136322 + }, + { + "epoch": 89, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.15308708516597747 + }, + { + "epoch": 90, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1539277541399002 + }, + { + "epoch": 91, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.15563526938915254 + }, + { + "epoch": 92, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15266645413398744 + }, + { + "epoch": 93, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.15498985040187835 + }, + { + "epoch": 94, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14707652615547182 + }, + { + "epoch": 95, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15589172424316405 + }, + { + "epoch": 96, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1510642097234726 + }, + { + "epoch": 97, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1477131044793129 + }, + { + "epoch": 98, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1512059341239929 + }, + { + "epoch": 99, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1547452706193924 + }, + { + "epoch": 100, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14908998366594314 + }, + { + "epoch": 101, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08004976162791252 + }, + { + "epoch": 102, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05500232973217964 + }, + { + "epoch": 103, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04631398814439774 + }, + { + "epoch": 104, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04183367156863212 + }, + { + "epoch": 105, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03675910843372345 + }, + { + "epoch": 106, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.03236614109992981 + }, + { + "epoch": 107, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.029470725601911545 + }, + { + "epoch": 108, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.026720388967990876 + }, + { + "epoch": 109, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.02745675653517246 + }, + { + "epoch": 110, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.024266670229136943 + }, + { + "epoch": 111, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02235375830113888 + }, + { + "epoch": 112, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.021399842903018 + }, + { + "epoch": 113, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.019970357822179794 + }, + { + "epoch": 114, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.018387679604291916 + }, + { + "epoch": 115, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.01710965284049511 + }, + { + "epoch": 116, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.016227416690587998 + }, + { + "epoch": 117, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01590447195649147 + }, + { + "epoch": 118, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.015522427982091904 + }, + { + "epoch": 119, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.013863605805635452 + }, + { + "epoch": 120, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.014408985558599234 + }, + { + "epoch": 121, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.013013962758034467 + }, + { + "epoch": 122, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01262190055757761 + }, + { + "epoch": 123, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.011193660405874252 + }, + { + "epoch": 124, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.012185472703874111 + }, + { + "epoch": 125, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.012042651894539595 + }, + { + "epoch": 126, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01146743201315403 + }, + { + "epoch": 127, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.011375230247974396 + }, + { + "epoch": 128, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.010315139727853239 + }, + { + "epoch": 129, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.009391635978072882 + }, + { + "epoch": 130, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009801293014287948 + }, + { + "epoch": 131, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.008947621556669474 + }, + { + "epoch": 132, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.008424072011113167 + }, + { + "epoch": 133, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.008684965382665395 + }, + { + "epoch": 134, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.00922296491727233 + }, + { + "epoch": 135, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.008562647834569215 + }, + { + "epoch": 136, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008652239821851253 + }, + { + "epoch": 137, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.008561775663793087 + }, + { + "epoch": 138, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.00765712965130806 + }, + { + "epoch": 139, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.007651031436175108 + }, + { + "epoch": 140, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.007763555864691734 + }, + { + "epoch": 141, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.007720564398616552 + }, + { + "epoch": 142, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.006966621245443821 + }, + { + "epoch": 143, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.007150115132853389 + }, + { + "epoch": 144, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007278579885214567 + }, + { + "epoch": 145, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.006301577565260232 + }, + { + "epoch": 146, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006381783106327057 + }, + { + "epoch": 147, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.00685056243032217 + }, + { + "epoch": 148, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006106687775440514 + }, + { + "epoch": 149, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.006147566365897655 + }, + { + "epoch": 150, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.006381408163979649 + }, + { + "epoch": 151, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005402126539442689 + }, + { + "epoch": 152, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.005967684311792254 + }, + { + "epoch": 153, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.004802315729185939 + }, + { + "epoch": 154, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004821319012865424 + }, + { + "epoch": 155, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.005573809953629971 + }, + { + "epoch": 156, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004940722731947899 + }, + { + "epoch": 157, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004846924012377858 + }, + { + "epoch": 158, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.005168056910932064 + }, + { + "epoch": 159, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.005365388280507177 + }, + { + "epoch": 160, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.0044669982519745825 + }, + { + "epoch": 161, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004651750999689102 + }, + { + "epoch": 162, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.0047598918889090415 + }, + { + "epoch": 163, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.004254822599813342 + }, + { + "epoch": 164, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.004679340793509036 + }, + { + "epoch": 165, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0050124835691601035 + }, + { + "epoch": 166, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.004335557008348406 + }, + { + "epoch": 167, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.004256838730871677 + }, + { + "epoch": 168, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.0041528904421255 + }, + { + "epoch": 169, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.004626856329888105 + }, + { + "epoch": 170, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.004153371350318193 + }, + { + "epoch": 171, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.004468166066259146 + }, + { + "epoch": 172, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.004028440014198422 + }, + { + "epoch": 173, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.004062429394945502 + }, + { + "epoch": 174, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0044060234404355286 + }, + { + "epoch": 175, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.004328454611673951 + }, + { + "epoch": 176, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.004056132428869605 + }, + { + "epoch": 177, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.004445703057423234 + }, + { + "epoch": 178, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.003961552263256163 + }, + { + "epoch": 179, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0038441047351807354 + }, + { + "epoch": 180, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.00424440163590014 + }, + { + "epoch": 181, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0037770761930942536 + }, + { + "epoch": 182, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.003911129658930004 + }, + { + "epoch": 183, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.003980180521793664 + }, + { + "epoch": 184, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.004038504354581237 + }, + { + "epoch": 185, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.004582509907409549 + }, + { + "epoch": 186, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.003771644706055522 + }, + { + "epoch": 187, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.003987695770338178 + }, + { + "epoch": 188, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0038091223045438527 + }, + { + "epoch": 189, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003995611964538693 + }, + { + "epoch": 190, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.004000969804227352 + }, + { + "epoch": 191, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0037076463522017003 + }, + { + "epoch": 192, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0034565704812482 + }, + { + "epoch": 193, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.003468065955489874 + }, + { + "epoch": 194, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.004178308155275881 + }, + { + "epoch": 195, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0038809054918959736 + }, + { + "epoch": 196, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.003907086752802134 + }, + { + "epoch": 197, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.004144384138062596 + }, + { + "epoch": 198, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0036071151252090932 + }, + { + "epoch": 199, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0033969543035514654 + }, + { + "epoch": 200, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.003727299030162394 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9251, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.371705908203125, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3320109568, + "peak_memory_reserved_bytes": 4542431232, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.065307855606079, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10386.89548611641, + "train_wall_s": 10384.79561829567 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2509713920000000, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 0, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 3764570880000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json new file mode 100644 index 0000000..ee6a66c --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json @@ -0,0 +1,3390 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9996285140514374, + "feedback_forward_cosine": [ + 0.9996448755264282, + 0.9996241927146912, + 0.9996662139892578, + 0.9996505379676819, + 0.9997282028198242, + 0.9997563362121582, + 0.9997356534004211, + 0.9996742010116577, + 0.9997925162315369, + 0.9997165203094482, + 0.9997353553771973, + 0.9998100996017456, + 0.9997087717056274, + 0.9996922016143799, + 0.999758243560791, + 0.9997119903564453, + 0.9997909665107727, + 0.9997221231460571, + 0.9998180866241455, + 0.999744713306427, + 0.9998014569282532, + 0.9998766183853149, + 0.9998487830162048, + 0.9998272657394409, + 0.9998286962509155, + 0.9997783899307251, + 0.9998400211334229, + 0.9997010231018066, + 0.9996979832649231, + 0.9994108080863953, + 0.9999839067459106 + ], + "feedback_forward_norm_ratio": [ + 1.0001192092895508, + 1.0009658336639404, + 0.9994824528694153, + 1.0007823705673218, + 0.9986081719398499, + 1.0005202293395996, + 0.9998824000358582, + 1.0015058517456055, + 0.9998427033424377, + 1.0006579160690308, + 0.9985700249671936, + 1.0006816387176514, + 0.9992146492004395, + 1.0012266635894775, + 0.9997763633728027, + 1.0012537240982056, + 0.9999839663505554, + 1.0013760328292847, + 0.9994275569915771, + 1.0008691549301147, + 0.9993569254875183, + 1.0003502368927002, + 0.9998308420181274, + 1.0005744695663452, + 0.9997363090515137, + 1.0010242462158203, + 0.9999058842658997, + 1.0010095834732056, + 0.9994451999664307, + 1.0014487504959106, + 1.0020092725753784 + ], + "feedback_forward_relative_error": [ + 0.026650002226233482, + 0.027447007596492767, + 0.025834619998931885, + 0.0264601930975914, + 0.02333914116024971, + 0.02209114469587803, + 0.02299085073173046, + 0.02558811567723751, + 0.020373323932290077, + 0.023823890835046768, + 0.023036517202854156, + 0.019504692405462265, + 0.02413417026400566, + 0.02485603094100952, + 0.02198750525712967, + 0.024049103260040283, + 0.020448429509997368, + 0.02362847700715065, + 0.019075661897659302, + 0.022619416937232018, + 0.01992790214717388, + 0.01571011357009411, + 0.017386458814144135, + 0.018599554896354675, + 0.018510015681385994, + 0.021088147535920143, + 0.017884954810142517, + 0.024482930079102516, + 0.024578392505645752, + 0.034382302314043045, + 0.006041166838258505 + ], + "innovation_negative_gradient_cosine": [ + 0.9996632933616638, + 0.9995239973068237, + 0.9996660947799683, + 0.9995747804641724, + 0.9996673464775085, + 0.9996167421340942, + 0.9996652007102966, + 0.9995874166488647, + 0.9996679425239563, + 0.9996523261070251, + 0.9996814727783203, + 0.9996957778930664, + 0.9996981620788574, + 0.9996000528335571, + 0.9997052550315857, + 0.9996254444122314, + 0.9997178316116333, + 0.9996660947799683, + 0.9997333884239197, + 0.9997081756591797, + 0.9997397065162659, + 0.9997773766517639, + 0.9997949004173279, + 0.9997736215591431, + 0.9997889995574951, + 0.99980628490448, + 0.9998396635055542, + 0.9998664259910583, + 0.9998698234558105, + 0.9999217987060547, + 0.7832923531532288 + ], + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.02214613657504801, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9996632933616638, + 0.9995239973068237, + 0.9996660947799683, + 0.9995747804641724, + 0.9996673464775085, + 0.9996167421340942, + 0.9996652007102966, + 0.9995874166488647, + 0.9996679425239563, + 0.9996523261070251, + 0.9996814727783203, + 0.9996957778930664, + 0.9996981620788574, + 0.9996000528335571, + 0.9997052550315857, + 0.9996254444122314, + 0.9997178316116333, + 0.9996660947799683, + 0.9997333884239197, + 0.9997081756591797, + 0.9997397065162659, + 0.9997773766517639, + 0.9997949004173279, + 0.9997736215591431, + 0.9997889995574951, + 0.99980628490448, + 0.9998396635055542, + 0.9998664259910583, + 0.9998698234558105, + 0.9999217987060547, + 0.7832923531532288 + ], + "teaching_negative_gradient_cosine": [ + 0.9996632933616638, + 0.9995239973068237, + 0.9996660947799683, + 0.9995747804641724, + 0.9996673464775085, + 0.9996167421340942, + 0.9996652007102966, + 0.9995874166488647, + 0.9996679425239563, + 0.9996523261070251, + 0.9996814727783203, + 0.9996957778930664, + 0.9996981620788574, + 0.9996000528335571, + 0.9997052550315857, + 0.9996254444122314, + 0.9997178316116333, + 0.9996660947799683, + 0.9997333884239197, + 0.9997081756591797, + 0.9997397065162659, + 0.9997773766517639, + 0.9997949004173279, + 0.9997736215591431, + 0.9997889995574951, + 0.99980628490448, + 0.9998396635055542, + 0.9998664259910583, + 0.9998698234558105, + 0.9999217987060547, + 0.7832923531532288 + ], + "wall_s": 0.13054966926574707 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3585952520370483, + "mean_feedback_forward_cosine": 0.3046831685689188, + "mean_feedback_forward_relative_error": 1.1702621463806397, + "min_feedback_forward_cosine": 0.08090806752443314 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.1066284007263185 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3304003477096558, + "mean_feedback_forward_cosine": 0.36645406773013456, + "mean_feedback_forward_relative_error": 1.1181111816437013, + "min_feedback_forward_cosine": 0.11908140778541565 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.508727656364441 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.299444317817688, + "mean_feedback_forward_cosine": 0.4346453778205379, + "mean_feedback_forward_relative_error": 1.057305653249064, + "min_feedback_forward_cosine": 0.16057761013507843 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.2519019562149047 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2654600143432617, + "mean_feedback_forward_cosine": 0.4959667310599358, + "mean_feedback_forward_relative_error": 0.9992632356382185, + "min_feedback_forward_cosine": 0.20524783432483673 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.0251145127868653 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2304916381835938, + "mean_feedback_forward_cosine": 0.5455621259827768, + "mean_feedback_forward_relative_error": 0.9495349939792387, + "min_feedback_forward_cosine": 0.2493714690208435 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8638715417480469 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.199565052986145, + "mean_feedback_forward_cosine": 0.5886594960766454, + "mean_feedback_forward_relative_error": 0.9039133075744875, + "min_feedback_forward_cosine": 0.29433685541152954 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7554565207672119 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.171512484550476, + "mean_feedback_forward_cosine": 0.6259271387130984, + "mean_feedback_forward_relative_error": 0.8622603945193752, + "min_feedback_forward_cosine": 0.33613449335098267 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6826584720993042 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1432757377624512, + "mean_feedback_forward_cosine": 0.6577128892944705, + "mean_feedback_forward_relative_error": 0.8246829057893446, + "min_feedback_forward_cosine": 0.3702732026576996 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6287693549728394 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1165492534637451, + "mean_feedback_forward_cosine": 0.685499252811555, + "mean_feedback_forward_relative_error": 0.7903105520432995, + "min_feedback_forward_cosine": 0.40048497915267944 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5804982432174682 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0891292095184326, + "mean_feedback_forward_cosine": 0.7110802377423933, + "mean_feedback_forward_relative_error": 0.7574431088662916, + "min_feedback_forward_cosine": 0.43083637952804565 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5510316325569152 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0645053386688232, + "mean_feedback_forward_cosine": 0.7334197105899933, + "mean_feedback_forward_relative_error": 0.7274134264838311, + "min_feedback_forward_cosine": 0.45720577239990234 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5122311639595032 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0398348569869995, + "mean_feedback_forward_cosine": 0.753688100845583, + "mean_feedback_forward_relative_error": 0.6989298068707989, + "min_feedback_forward_cosine": 0.4827224612236023 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.49010992847442625 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0150221586227417, + "mean_feedback_forward_cosine": 0.7723052020995848, + "mean_feedback_forward_relative_error": 0.6717068211686227, + "min_feedback_forward_cosine": 0.5074095726013184 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4641452179527283 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9889848232269287, + "mean_feedback_forward_cosine": 0.7893762050136444, + "mean_feedback_forward_relative_error": 0.6456553873515898, + "min_feedback_forward_cosine": 0.5324095487594604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4428069670677185 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9631386995315552, + "mean_feedback_forward_cosine": 0.8049058818048046, + "mean_feedback_forward_relative_error": 0.6210459845681344, + "min_feedback_forward_cosine": 0.5565786957740784 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4246428520202637 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.937900722026825, + "mean_feedback_forward_cosine": 0.8192884499026883, + "mean_feedback_forward_relative_error": 0.5973106839003102, + "min_feedback_forward_cosine": 0.5796732902526855 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4130655108165741 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9116066694259644, + "mean_feedback_forward_cosine": 0.8326400441508139, + "mean_feedback_forward_relative_error": 0.5743048070899902, + "min_feedback_forward_cosine": 0.6022407412528992 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.39465926347732544 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8865914344787598, + "mean_feedback_forward_cosine": 0.8449586822140601, + "mean_feedback_forward_relative_error": 0.5525295075870329, + "min_feedback_forward_cosine": 0.6241908073425293 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.38214210516929625 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8610144257545471, + "mean_feedback_forward_cosine": 0.856250032301872, + "mean_feedback_forward_relative_error": 0.5316746835747073, + "min_feedback_forward_cosine": 0.6453561186790466 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.37219429008483884 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8360261917114258, + "mean_feedback_forward_cosine": 0.8666744674405744, + "mean_feedback_forward_relative_error": 0.5117127260854167, + "min_feedback_forward_cosine": 0.6658006906509399 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.355763973941803 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8105968832969666, + "mean_feedback_forward_cosine": 0.8765667311606868, + "mean_feedback_forward_relative_error": 0.4920242529722952, + "min_feedback_forward_cosine": 0.6857298016548157 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3471485916996002 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7856993675231934, + "mean_feedback_forward_cosine": 0.8855374013223956, + "mean_feedback_forward_relative_error": 0.4736229320687632, + "min_feedback_forward_cosine": 0.7045443654060364 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3388717658805847 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.761393666267395, + "mean_feedback_forward_cosine": 0.893738733183953, + "mean_feedback_forward_relative_error": 0.45595797127293003, + "min_feedback_forward_cosine": 0.7220780253410339 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.32842747619628904 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7373024821281433, + "mean_feedback_forward_cosine": 0.9013767088613203, + "mean_feedback_forward_relative_error": 0.4389302769014912, + "min_feedback_forward_cosine": 0.7390848398208618 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.31618609956741334 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7132046818733215, + "mean_feedback_forward_cosine": 0.9085880940960299, + "mean_feedback_forward_relative_error": 0.4223930532893827, + "min_feedback_forward_cosine": 0.7556705474853516 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31259035013198855 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6883693337440491, + "mean_feedback_forward_cosine": 0.9153294005701619, + "mean_feedback_forward_relative_error": 0.4062719056683202, + "min_feedback_forward_cosine": 0.7722120881080627 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31096653012275693 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6657470464706421, + "mean_feedback_forward_cosine": 0.9213306019383092, + "mean_feedback_forward_relative_error": 0.3914064000691137, + "min_feedback_forward_cosine": 0.7868608236312866 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2952206775283813 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6425599455833435, + "mean_feedback_forward_cosine": 0.9272541884453066, + "mean_feedback_forward_relative_error": 0.3762246250144897, + "min_feedback_forward_cosine": 0.8012421131134033 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2912610262298584 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6199831366539001, + "mean_feedback_forward_cosine": 0.9326016787559755, + "mean_feedback_forward_relative_error": 0.36191003024578094, + "min_feedback_forward_cosine": 0.8147948384284973 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28817487318992613 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.597847044467926, + "mean_feedback_forward_cosine": 0.9374953354558637, + "mean_feedback_forward_relative_error": 0.3483095572840783, + "min_feedback_forward_cosine": 0.8276832699775696 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2810961621761322 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5762627124786377, + "mean_feedback_forward_cosine": 0.9419867396354675, + "mean_feedback_forward_relative_error": 0.3354184043503577, + "min_feedback_forward_cosine": 0.8397367596626282 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27481924999713897 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.554566502571106, + "mean_feedback_forward_cosine": 0.9463227083606105, + "mean_feedback_forward_relative_error": 0.32253597099934855, + "min_feedback_forward_cosine": 0.8514574766159058 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2712509515380859 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5340067744255066, + "mean_feedback_forward_cosine": 0.9503983489928707, + "mean_feedback_forward_relative_error": 0.3099248527038482, + "min_feedback_forward_cosine": 0.8621289730072021 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2722747924566269 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5158315300941467, + "mean_feedback_forward_cosine": 0.9539481401443481, + "mean_feedback_forward_relative_error": 0.2984740154877786, + "min_feedback_forward_cosine": 0.8712555170059204 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25798445239067075 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.496427059173584, + "mean_feedback_forward_cosine": 0.9573978377926734, + "mean_feedback_forward_relative_error": 0.28698549996460637, + "min_feedback_forward_cosine": 0.8806583881378174 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2554719933319092 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4773749113082886, + "mean_feedback_forward_cosine": 0.9606026795602614, + "mean_feedback_forward_relative_error": 0.27592927049244603, + "min_feedback_forward_cosine": 0.8895782232284546 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25545837335586546 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4594625234603882, + "mean_feedback_forward_cosine": 0.9635426498228504, + "mean_feedback_forward_relative_error": 0.26538106630886754, + "min_feedback_forward_cosine": 0.8976370692253113 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2484088640499115 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4422893226146698, + "mean_feedback_forward_cosine": 0.9661906438489114, + "mean_feedback_forward_relative_error": 0.25547523916729037, + "min_feedback_forward_cosine": 0.9050342440605164 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2440171889591217 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42448484897613525, + "mean_feedback_forward_cosine": 0.9687469697767689, + "mean_feedback_forward_relative_error": 0.24559208870895446, + "min_feedback_forward_cosine": 0.912443220615387 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.24323479767799377 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4080718159675598, + "mean_feedback_forward_cosine": 0.9710481493703781, + "mean_feedback_forward_relative_error": 0.23629709813864, + "min_feedback_forward_cosine": 0.9190430641174316 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24011920569419862 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3921796679496765, + "mean_feedback_forward_cosine": 0.9732197542344371, + "mean_feedback_forward_relative_error": 0.22721118720308428, + "min_feedback_forward_cosine": 0.9250966906547546 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2373897479915619 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3771727383136749, + "mean_feedback_forward_cosine": 0.9752226568037464, + "mean_feedback_forward_relative_error": 0.21850634558546927, + "min_feedback_forward_cosine": 0.9306812882423401 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23271506587028504 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36220383644104004, + "mean_feedback_forward_cosine": 0.9770623618556608, + "mean_feedback_forward_relative_error": 0.21018394178921176, + "min_feedback_forward_cosine": 0.9360086917877197 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.23216551981925965 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34794294834136963, + "mean_feedback_forward_cosine": 0.9787843784978313, + "mean_feedback_forward_relative_error": 0.20208276856330135, + "min_feedback_forward_cosine": 0.9409137964248657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.23053674859046935 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33451753854751587, + "mean_feedback_forward_cosine": 0.9803864609810614, + "mean_feedback_forward_relative_error": 0.1942895285544857, + "min_feedback_forward_cosine": 0.9453545808792114 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2259896070957184 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32157811522483826, + "mean_feedback_forward_cosine": 0.9818134846225861, + "mean_feedback_forward_relative_error": 0.1870680811905092, + "min_feedback_forward_cosine": 0.9494665265083313 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2214436098384857 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3088374733924866, + "mean_feedback_forward_cosine": 0.9831757333970839, + "mean_feedback_forward_relative_error": 0.17989249154925346, + "min_feedback_forward_cosine": 0.9533370733261108 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21873262013435363 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29687994718551636, + "mean_feedback_forward_cosine": 0.9844553701339229, + "mean_feedback_forward_relative_error": 0.17289376955839894, + "min_feedback_forward_cosine": 0.9568438529968262 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21603742006778717 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28457173705101013, + "mean_feedback_forward_cosine": 0.9856576804191836, + "mean_feedback_forward_relative_error": 0.1660312604038946, + "min_feedback_forward_cosine": 0.9602963924407959 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21823587814807893 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27344414591789246, + "mean_feedback_forward_cosine": 0.9867185546505836, + "mean_feedback_forward_relative_error": 0.1597499662349301, + "min_feedback_forward_cosine": 0.9633457660675049 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21540137950897217 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26242485642433167, + "mean_feedback_forward_cosine": 0.9877236908481967, + "mean_feedback_forward_relative_error": 0.15357495339647417, + "min_feedback_forward_cosine": 0.9662054777145386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.21133698526382447 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25207939743995667, + "mean_feedback_forward_cosine": 0.9886252514777645, + "mean_feedback_forward_relative_error": 0.14780485077250388, + "min_feedback_forward_cosine": 0.9687892198562622 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2037308853340149 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24176177382469177, + "mean_feedback_forward_cosine": 0.9894983691553916, + "mean_feedback_forward_relative_error": 0.1420233254471133, + "min_feedback_forward_cosine": 0.9712689518928528 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.21174614361763 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23217424750328064, + "mean_feedback_forward_cosine": 0.9902935451076876, + "mean_feedback_forward_relative_error": 0.13653598281164323, + "min_feedback_forward_cosine": 0.97348552942276 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.20778529538154603 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22279824316501617, + "mean_feedback_forward_cosine": 0.9910120848686464, + "mean_feedback_forward_relative_error": 0.13139018248165807, + "min_feedback_forward_cosine": 0.9755691885948181 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.20543151748657226 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21348775923252106, + "mean_feedback_forward_cosine": 0.9916909317816457, + "mean_feedback_forward_relative_error": 0.12633225069411338, + "min_feedback_forward_cosine": 0.9775524139404297 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.20323132181167602 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20458975434303284, + "mean_feedback_forward_cosine": 0.9923476730623553, + "mean_feedback_forward_relative_error": 0.12124136775251358, + "min_feedback_forward_cosine": 0.9793746471405029 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20717589849948884 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19673946499824524, + "mean_feedback_forward_cosine": 0.9928949167651515, + "mean_feedback_forward_relative_error": 0.11681427746530502, + "min_feedback_forward_cosine": 0.9809081554412842 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19442318168640138 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18847723305225372, + "mean_feedback_forward_cosine": 0.9934470076714793, + "mean_feedback_forward_relative_error": 0.11220145477883277, + "min_feedback_forward_cosine": 0.9824700355529785 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.2030546386051178 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18114864826202393, + "mean_feedback_forward_cosine": 0.9939348832253487, + "mean_feedback_forward_relative_error": 0.10793772243684338, + "min_feedback_forward_cosine": 0.9837960600852966 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19493575808525085 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17370857298374176, + "mean_feedback_forward_cosine": 0.9943953079562033, + "mean_feedback_forward_relative_error": 0.10375529179169286, + "min_feedback_forward_cosine": 0.9850887060165405 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19727207220554352 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.166924849152565, + "mean_feedback_forward_cosine": 0.9948073298700394, + "mean_feedback_forward_relative_error": 0.09986100598208365, + "min_feedback_forward_cosine": 0.9862250089645386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18774492676734925 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16006633639335632, + "mean_feedback_forward_cosine": 0.9952024394466031, + "mean_feedback_forward_relative_error": 0.09599625006798775, + "min_feedback_forward_cosine": 0.9873306751251221 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.19298322952270508 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15336787700653076, + "mean_feedback_forward_cosine": 0.9955765701109364, + "mean_feedback_forward_relative_error": 0.09218034304438098, + "min_feedback_forward_cosine": 0.9883619546890259 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.19527318564414978 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1468193382024765, + "mean_feedback_forward_cosine": 0.9959285663020226, + "mean_feedback_forward_relative_error": 0.08844295378413893, + "min_feedback_forward_cosine": 0.9893286824226379 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1948135542345047 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14080898463726044, + "mean_feedback_forward_cosine": 0.9962240880535495, + "mean_feedback_forward_relative_error": 0.08517507415625357, + "min_feedback_forward_cosine": 0.9901782274246216 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1880095457839966 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1351540982723236, + "mean_feedback_forward_cosine": 0.9965042529567596, + "mean_feedback_forward_relative_error": 0.08195698453534034, + "min_feedback_forward_cosine": 0.990946888923645 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18653369692325592 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1296241581439972, + "mean_feedback_forward_cosine": 0.9967624025960122, + "mean_feedback_forward_relative_error": 0.07887501467860514, + "min_feedback_forward_cosine": 0.9916706681251526 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18662398671150207 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12416495382785797, + "mean_feedback_forward_cosine": 0.9970199235023991, + "mean_feedback_forward_relative_error": 0.07567659170637207, + "min_feedback_forward_cosine": 0.9923562407493591 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.19312430601119995 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11924664676189423, + "mean_feedback_forward_cosine": 0.9972401665103051, + "mean_feedback_forward_relative_error": 0.07282290667776138, + "min_feedback_forward_cosine": 0.9929476976394653 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.18571868690490723 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11436894536018372, + "mean_feedback_forward_cosine": 0.9974491730813058, + "mean_feedback_forward_relative_error": 0.07000911301903186, + "min_feedback_forward_cosine": 0.9935097694396973 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18454137994766234 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10968337208032608, + "mean_feedback_forward_cosine": 0.9976482679766994, + "mean_feedback_forward_relative_error": 0.06722739471062537, + "min_feedback_forward_cosine": 0.9940293431282043 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.189456130194664 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10538037121295929, + "mean_feedback_forward_cosine": 0.9978226711673122, + "mean_feedback_forward_relative_error": 0.06468745921888659, + "min_feedback_forward_cosine": 0.9944861531257629 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18165870635032655 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10099851340055466, + "mean_feedback_forward_cosine": 0.9979852380291108, + "mean_feedback_forward_relative_error": 0.06222016592660258, + "min_feedback_forward_cosine": 0.9949323534965515 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17878232320785523 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09699861705303192, + "mean_feedback_forward_cosine": 0.9981338477903797, + "mean_feedback_forward_relative_error": 0.05988543576771213, + "min_feedback_forward_cosine": 0.9953258037567139 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17694817928314208 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09301231056451797, + "mean_feedback_forward_cosine": 0.998277444993296, + "mean_feedback_forward_relative_error": 0.057541606046499744, + "min_feedback_forward_cosine": 0.9957007765769958 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.18061958540916442 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08925621211528778, + "mean_feedback_forward_cosine": 0.9984065351947662, + "mean_feedback_forward_relative_error": 0.05534977417799734, + "min_feedback_forward_cosine": 0.9960404634475708 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17779186904907227 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08562029153108597, + "mean_feedback_forward_cosine": 0.9985263193807294, + "mean_feedback_forward_relative_error": 0.05322876013815403, + "min_feedback_forward_cosine": 0.9963556528091431 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17973715352058411 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08209428936243057, + "mean_feedback_forward_cosine": 0.9986398066243818, + "mean_feedback_forward_relative_error": 0.051138524807268576, + "min_feedback_forward_cosine": 0.9966488480567932 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17942093858718872 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07883550971746445, + "mean_feedback_forward_cosine": 0.998742280467864, + "mean_feedback_forward_relative_error": 0.049174504955449415, + "min_feedback_forward_cosine": 0.9969099760055542 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1764600699710846 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07566260546445847, + "mean_feedback_forward_cosine": 0.9988359059056928, + "mean_feedback_forward_relative_error": 0.04730717038675662, + "min_feedback_forward_cosine": 0.99715256690979 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17309480888843537 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07258174568414688, + "mean_feedback_forward_cosine": 0.9989236785519507, + "mean_feedback_forward_relative_error": 0.04549225762246117, + "min_feedback_forward_cosine": 0.997380256652832 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.17379588204622268 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06965117156505585, + "mean_feedback_forward_cosine": 0.9990047389461149, + "mean_feedback_forward_relative_error": 0.043746734849147255, + "min_feedback_forward_cosine": 0.9975866079330444 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.17552788778781891 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06684258580207825, + "mean_feedback_forward_cosine": 0.9990783602960648, + "mean_feedback_forward_relative_error": 0.04210145669358392, + "min_feedback_forward_cosine": 0.997776985168457 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1720303683567047 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06409264355897903, + "mean_feedback_forward_cosine": 0.9991503915479106, + "mean_feedback_forward_relative_error": 0.040425917464158224, + "min_feedback_forward_cosine": 0.9979556202888489 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.1756158047270775 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0615408718585968, + "mean_feedback_forward_cosine": 0.9992139647083897, + "mean_feedback_forward_relative_error": 0.0388862966229358, + "min_feedback_forward_cosine": 0.9981147646903992 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16958078844070434 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05904845520853996, + "mean_feedback_forward_cosine": 0.99927399235387, + "mean_feedback_forward_relative_error": 0.03737387319485987, + "min_feedback_forward_cosine": 0.9982644319534302 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1693227526807785 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05658656731247902, + "mean_feedback_forward_cosine": 0.9993299745744274, + "mean_feedback_forward_relative_error": 0.03590853398125018, + "min_feedback_forward_cosine": 0.9984058737754822 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.17388515575170516 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05434687063097954, + "mean_feedback_forward_cosine": 0.9993810269140428, + "mean_feedback_forward_relative_error": 0.03451352797809147, + "min_feedback_forward_cosine": 0.9985294938087463 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16667771428108216 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05209857597947121, + "mean_feedback_forward_cosine": 0.9994277108100152, + "mean_feedback_forward_relative_error": 0.03318805929513708, + "min_feedback_forward_cosine": 0.9986482858657837 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16874846799850463 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04997039586305618, + "mean_feedback_forward_cosine": 0.9994707511317346, + "mean_feedback_forward_relative_error": 0.03191607415435776, + "min_feedback_forward_cosine": 0.9987563490867615 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.17165495090007782 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04805481806397438, + "mean_feedback_forward_cosine": 0.9995101863338102, + "mean_feedback_forward_relative_error": 0.030704457402950334, + "min_feedback_forward_cosine": 0.9988495111465454 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16439861055374144 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04608337581157684, + "mean_feedback_forward_cosine": 0.9995480871969654, + "mean_feedback_forward_relative_error": 0.029492572520769412, + "min_feedback_forward_cosine": 0.9989418983459473 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16936889218330384 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.044296909123659134, + "mean_feedback_forward_cosine": 0.9995824367769303, + "mean_feedback_forward_relative_error": 0.028349369043304075, + "min_feedback_forward_cosine": 0.9990223050117493 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1659735676765442 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04257083311676979, + "mean_feedback_forward_cosine": 0.9996139830158602, + "mean_feedback_forward_relative_error": 0.027255154545268705, + "min_feedback_forward_cosine": 0.9990971088409424 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16362881194591522 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04086729884147644, + "mean_feedback_forward_cosine": 0.9996434719331803, + "mean_feedback_forward_relative_error": 0.026196662096246597, + "min_feedback_forward_cosine": 0.9991677403450012 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1682988141012192 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039153311401605606, + "mean_feedback_forward_cosine": 0.9996712015521142, + "mean_feedback_forward_relative_error": 0.025161175898486567, + "min_feedback_forward_cosine": 0.999235987663269 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.16988255014419557 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0376581996679306, + "mean_feedback_forward_cosine": 0.9996954664107291, + "mean_feedback_forward_relative_error": 0.024214518917424065, + "min_feedback_forward_cosine": 0.9992931485176086 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16026985026359558 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.036105673760175705, + "mean_feedback_forward_cosine": 0.999718306526061, + "mean_feedback_forward_relative_error": 0.023289367586614623, + "min_feedback_forward_cosine": 0.9993501901626587 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16757982426166534 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034648407250642776, + "mean_feedback_forward_cosine": 0.9997396526798126, + "mean_feedback_forward_relative_error": 0.022391358210194494, + "min_feedback_forward_cosine": 0.9994015693664551 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.16521677165985107 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034636080265045166, + "mean_feedback_forward_cosine": 0.9997398238028249, + "mean_feedback_forward_relative_error": 0.022381956839273052, + "min_feedback_forward_cosine": 0.9994021654129028 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09520915712833404 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0346299484372139, + "mean_feedback_forward_cosine": 0.9997399507030365, + "mean_feedback_forward_relative_error": 0.022375703503888462, + "min_feedback_forward_cosine": 0.9994022846221924 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06890638122200966 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034624312072992325, + "mean_feedback_forward_cosine": 0.9997400160758726, + "mean_feedback_forward_relative_error": 0.0223699755996706, + "min_feedback_forward_cosine": 0.9994024634361267 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.06154520476341248 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03461897745728493, + "mean_feedback_forward_cosine": 0.9997401256715098, + "mean_feedback_forward_relative_error": 0.02236448127716299, + "min_feedback_forward_cosine": 0.9994025230407715 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.05448996238946915 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03461356833577156, + "mean_feedback_forward_cosine": 0.9997402468035298, + "mean_feedback_forward_relative_error": 0.022359185491598422, + "min_feedback_forward_cosine": 0.9994028210639954 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04973544316411018 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034608371555805206, + "mean_feedback_forward_cosine": 0.9997403102536355, + "mean_feedback_forward_relative_error": 0.022353943377252546, + "min_feedback_forward_cosine": 0.9994028806686401 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.045988887414932254 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034603506326675415, + "mean_feedback_forward_cosine": 0.999740389085585, + "mean_feedback_forward_relative_error": 0.02234901749198475, + "min_feedback_forward_cosine": 0.9994032979011536 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.04122027599394321 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034598417580127716, + "mean_feedback_forward_cosine": 0.9997404756084565, + "mean_feedback_forward_relative_error": 0.022344095572348562, + "min_feedback_forward_cosine": 0.9994032382965088 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03964277899980545 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034593500196933746, + "mean_feedback_forward_cosine": 0.9997405717449803, + "mean_feedback_forward_relative_error": 0.022339207615943685, + "min_feedback_forward_cosine": 0.9994034767150879 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03581520953893662 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03458848223090172, + "mean_feedback_forward_cosine": 0.999740683263348, + "mean_feedback_forward_relative_error": 0.022334380976615414, + "min_feedback_forward_cosine": 0.9994037747383118 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.0351124484539032 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03458338975906372, + "mean_feedback_forward_cosine": 0.9997407447907233, + "mean_feedback_forward_relative_error": 0.022329576583879608, + "min_feedback_forward_cosine": 0.9994038343429565 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.034428428021669386 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034578461199998856, + "mean_feedback_forward_cosine": 0.9997408543863604, + "mean_feedback_forward_relative_error": 0.022324893022737196, + "min_feedback_forward_cosine": 0.9994041323661804 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.03098803948402405 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03457332402467728, + "mean_feedback_forward_cosine": 0.9997409409092318, + "mean_feedback_forward_relative_error": 0.022320216416471427, + "min_feedback_forward_cosine": 0.9994043111801147 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.029546013591885568 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034568533301353455, + "mean_feedback_forward_cosine": 0.9997410428139472, + "mean_feedback_forward_relative_error": 0.022315731632613366, + "min_feedback_forward_cosine": 0.9994044303894043 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.028411184762120248 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03456365689635277, + "mean_feedback_forward_cosine": 0.9997411178004357, + "mean_feedback_forward_relative_error": 0.022311257078282295, + "min_feedback_forward_cosine": 0.9994045495986938 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02664868856549263 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03455884009599686, + "mean_feedback_forward_cosine": 0.9997411927869243, + "mean_feedback_forward_relative_error": 0.022306785378004273, + "min_feedback_forward_cosine": 0.9994048476219177 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.026244018102288245 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03455414995551109, + "mean_feedback_forward_cosine": 0.9997412850779872, + "mean_feedback_forward_relative_error": 0.022302366192302397, + "min_feedback_forward_cosine": 0.9994049072265625 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.02401714214324951 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03454875573515892, + "mean_feedback_forward_cosine": 0.999741373523589, + "mean_feedback_forward_relative_error": 0.022297664245049798, + "min_feedback_forward_cosine": 0.9994050860404968 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.025217011668682097 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03454356640577316, + "mean_feedback_forward_cosine": 0.9997414523555387, + "mean_feedback_forward_relative_error": 0.02229296860675658, + "min_feedback_forward_cosine": 0.9994050860404968 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.0238461730080843 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034538786858320236, + "mean_feedback_forward_cosine": 0.999741550414793, + "mean_feedback_forward_relative_error": 0.022288713769446456, + "min_feedback_forward_cosine": 0.9994054436683655 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.021576540098786356 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03453381359577179, + "mean_feedback_forward_cosine": 0.9997416157876292, + "mean_feedback_forward_relative_error": 0.02228425509266315, + "min_feedback_forward_cosine": 0.9994056224822998 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.022200865411758423 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034529585391283035, + "mean_feedback_forward_cosine": 0.9997417292287273, + "mean_feedback_forward_relative_error": 0.022280036637018764, + "min_feedback_forward_cosine": 0.9994057416915894 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.019916686898469926 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03452460095286369, + "mean_feedback_forward_cosine": 0.9997417946015635, + "mean_feedback_forward_relative_error": 0.022275654343707908, + "min_feedback_forward_cosine": 0.9994059801101685 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.01987594199478626 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03451919928193092, + "mean_feedback_forward_cosine": 0.9997418984290092, + "mean_feedback_forward_relative_error": 0.022271059047911438, + "min_feedback_forward_cosine": 0.9994060397148132 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.020289315200299026 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034514255821704865, + "mean_feedback_forward_cosine": 0.9997419599563845, + "mean_feedback_forward_relative_error": 0.02226658763303872, + "min_feedback_forward_cosine": 0.9994062781333923 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.0189817143394053 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034508705139160156, + "mean_feedback_forward_cosine": 0.9997420676292912, + "mean_feedback_forward_relative_error": 0.022261766972200524, + "min_feedback_forward_cosine": 0.9994063973426819 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.020050239784419537 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03450360521674156, + "mean_feedback_forward_cosine": 0.9997421733794674, + "mean_feedback_forward_relative_error": 0.022257326411143426, + "min_feedback_forward_cosine": 0.9994065761566162 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.01772526730298996 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03449811786413193, + "mean_feedback_forward_cosine": 0.9997422695159912, + "mean_feedback_forward_relative_error": 0.0222526325153247, + "min_feedback_forward_cosine": 0.9994068741798401 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01855657254755497 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03449344262480736, + "mean_feedback_forward_cosine": 0.9997423541161322, + "mean_feedback_forward_relative_error": 0.022248169557461814, + "min_feedback_forward_cosine": 0.9994069337844849 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.01728311818420887 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03448819741606712, + "mean_feedback_forward_cosine": 0.9997424502526561, + "mean_feedback_forward_relative_error": 0.022243508783679818, + "min_feedback_forward_cosine": 0.999407172203064 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.017123812686502933 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03448304161429405, + "mean_feedback_forward_cosine": 0.9997425502346408, + "mean_feedback_forward_relative_error": 0.022238834580827142, + "min_feedback_forward_cosine": 0.9994072914123535 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.01670280195534229 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034477971494197845, + "mean_feedback_forward_cosine": 0.9997426271438599, + "mean_feedback_forward_relative_error": 0.02223430791749589, + "min_feedback_forward_cosine": 0.9994074702262878 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.01564242881298065 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03447258844971657, + "mean_feedback_forward_cosine": 0.999742736739497, + "mean_feedback_forward_relative_error": 0.022229604843643405, + "min_feedback_forward_cosine": 0.9994077682495117 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.015944313647747038 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034467972815036774, + "mean_feedback_forward_cosine": 0.9997428155714466, + "mean_feedback_forward_relative_error": 0.022225436378991412, + "min_feedback_forward_cosine": 0.9994078874588013 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.014028937001600862 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034463029354810715, + "mean_feedback_forward_cosine": 0.9997429078625094, + "mean_feedback_forward_relative_error": 0.022220811205765894, + "min_feedback_forward_cosine": 0.9994081258773804 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.014854636296331882 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03445770964026451, + "mean_feedback_forward_cosine": 0.999742982848998, + "mean_feedback_forward_relative_error": 0.02221618534155911, + "min_feedback_forward_cosine": 0.9994082450866699 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.01487931094944477 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034452635794878006, + "mean_feedback_forward_cosine": 0.9997430924446352, + "mean_feedback_forward_relative_error": 0.022211641779229525, + "min_feedback_forward_cosine": 0.9994083642959595 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.014188854694962501 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034447718411684036, + "mean_feedback_forward_cosine": 0.9997431905038895, + "mean_feedback_forward_relative_error": 0.022207115972114187, + "min_feedback_forward_cosine": 0.9994085431098938 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.013264953768849373 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03444201499223709, + "mean_feedback_forward_cosine": 0.9997433000995267, + "mean_feedback_forward_relative_error": 0.022202054278985146, + "min_feedback_forward_cosine": 0.9994086623191833 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.015078886508345603 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034436747431755066, + "mean_feedback_forward_cosine": 0.9997433731632848, + "mean_feedback_forward_relative_error": 0.02219739014042481, + "min_feedback_forward_cosine": 0.9994089603424072 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01349074018061161 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03443177789449692, + "mean_feedback_forward_cosine": 0.9997434654543477, + "mean_feedback_forward_relative_error": 0.022192851820540045, + "min_feedback_forward_cosine": 0.9994090795516968 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.0124130152207613 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03442655876278877, + "mean_feedback_forward_cosine": 0.9997435808181763, + "mean_feedback_forward_relative_error": 0.022188144811098615, + "min_feedback_forward_cosine": 0.9994092583656311 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.012687791421413422 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034421082586050034, + "mean_feedback_forward_cosine": 0.9997436827228915, + "mean_feedback_forward_relative_error": 0.022183355559865312, + "min_feedback_forward_cosine": 0.9994094371795654 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.012818007640689612 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034415971487760544, + "mean_feedback_forward_cosine": 0.9997438173140248, + "mean_feedback_forward_relative_error": 0.022178630044143045, + "min_feedback_forward_cosine": 0.9994096159934998 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.012154696775600314 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03441084176301956, + "mean_feedback_forward_cosine": 0.9997438807641307, + "mean_feedback_forward_relative_error": 0.02217404039636735, + "min_feedback_forward_cosine": 0.9994098544120789 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.011863310490995645 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03440537303686142, + "mean_feedback_forward_cosine": 0.9997439903597678, + "mean_feedback_forward_relative_error": 0.022169174971960245, + "min_feedback_forward_cosine": 0.9994099736213684 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.012201981163918973 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03439989686012268, + "mean_feedback_forward_cosine": 0.9997440807281002, + "mean_feedback_forward_relative_error": 0.02216440129784807, + "min_feedback_forward_cosine": 0.9994102120399475 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.011703400059193373 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034394387155771255, + "mean_feedback_forward_cosine": 0.9997441787873546, + "mean_feedback_forward_relative_error": 0.022159446226132494, + "min_feedback_forward_cosine": 0.9994103908538818 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.011733281977027655 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034389425069093704, + "mean_feedback_forward_cosine": 0.9997442749238783, + "mean_feedback_forward_relative_error": 0.02215483523304424, + "min_feedback_forward_cosine": 0.9994105100631714 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.011098120585232973 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034384094178676605, + "mean_feedback_forward_cosine": 0.9997443902877069, + "mean_feedback_forward_relative_error": 0.02215007922401832, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.010892319923341274 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034384019672870636, + "mean_feedback_forward_cosine": 0.999744378751324, + "mean_feedback_forward_relative_error": 0.022149969958850452, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008965941136330367 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438396751880646, + "mean_feedback_forward_cosine": 0.9997443922104374, + "mean_feedback_forward_relative_error": 0.022149880581925952, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.009145139462947845 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383926540613174, + "mean_feedback_forward_cosine": 0.999744386442246, + "mean_feedback_forward_relative_error": 0.022149790243636214, + "min_feedback_forward_cosine": 0.9994106888771057 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.00857024108633399 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438388183712959, + "mean_feedback_forward_cosine": 0.9997443845195155, + "mean_feedback_forward_relative_error": 0.022149701377437, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.00860008357539773 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438383713364601, + "mean_feedback_forward_cosine": 0.9997443768285936, + "mean_feedback_forward_relative_error": 0.022149615951122775, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.008307939018383621 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383803606033325, + "mean_feedback_forward_cosine": 0.9997443768285936, + "mean_feedback_forward_relative_error": 0.022149527475478187, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.008682450590580702 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438376262784004, + "mean_feedback_forward_cosine": 0.9997443941331678, + "mean_feedback_forward_relative_error": 0.022149444407513066, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.008061909982338548 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438372537493706, + "mean_feedback_forward_cosine": 0.9997444037468203, + "mean_feedback_forward_relative_error": 0.022149356923276377, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.008938052902668714 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438368812203407, + "mean_feedback_forward_cosine": 0.9997443922104374, + "mean_feedback_forward_relative_error": 0.022149273209393985, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.00822726613894105 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438365086913109, + "mean_feedback_forward_cosine": 0.9997443922104374, + "mean_feedback_forward_relative_error": 0.022149191207943425, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007490905041694641 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0343836210668087, + "mean_feedback_forward_cosine": 0.9997444056695507, + "mean_feedback_forward_relative_error": 0.02214910886100223, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.00793412824228406 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438359126448631, + "mean_feedback_forward_cosine": 0.9997443960558984, + "mean_feedback_forward_relative_error": 0.022149030990417925, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007137769331932068 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438355401158333, + "mean_feedback_forward_cosine": 0.9997444037468203, + "mean_feedback_forward_relative_error": 0.022148951332295132, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007553048278689385 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383516758680344, + "mean_feedback_forward_cosine": 0.9997443922104374, + "mean_feedback_forward_relative_error": 0.022148873446689497, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007888392472825944 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438347950577736, + "mean_feedback_forward_cosine": 0.9997443960558984, + "mean_feedback_forward_relative_error": 0.02214879309758544, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.007774316803961992 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438344597816467, + "mean_feedback_forward_cosine": 0.9997444037468203, + "mean_feedback_forward_relative_error": 0.02214871615832371, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.007526577415838838 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438340499997139, + "mean_feedback_forward_cosine": 0.9997443979786288, + "mean_feedback_forward_relative_error": 0.02214863417189448, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007590677718669176 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383371472358704, + "mean_feedback_forward_cosine": 0.999744386442246, + "mean_feedback_forward_relative_error": 0.022148555129646294, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.007188964843824506 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438333794474602, + "mean_feedback_forward_cosine": 0.9997444152832031, + "mean_feedback_forward_relative_error": 0.02214847773974461, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006880649688988924 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438330069184303, + "mean_feedback_forward_cosine": 0.9997443922104374, + "mean_feedback_forward_relative_error": 0.02214839567820872, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.007505728708356619 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438326343894005, + "mean_feedback_forward_cosine": 0.9997444056695507, + "mean_feedback_forward_relative_error": 0.022148319580141577, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0073278361514955755 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438323363661766, + "mean_feedback_forward_cosine": 0.9997444095150116, + "mean_feedback_forward_relative_error": 0.02214824267092251, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006741334313601255 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438319265842438, + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.022148165776724776, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0072374745182693006 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438316285610199, + "mean_feedback_forward_cosine": 0.9997444018240897, + "mean_feedback_forward_relative_error": 0.02214809227734804, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.007213199753910303 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0343831330537796, + "mean_feedback_forward_cosine": 0.9997444075922812, + "mean_feedback_forward_relative_error": 0.02214801611919557, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0068370764482021335 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383099526166916, + "mean_feedback_forward_cosine": 0.9997443999013593, + "mean_feedback_forward_relative_error": 0.022147933141358437, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.007554607146382332 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438306227326393, + "mean_feedback_forward_cosine": 0.9997444037468203, + "mean_feedback_forward_relative_error": 0.022147857238568606, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.007335109049528837 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034383028745651245, + "mean_feedback_forward_cosine": 0.9997444037468203, + "mean_feedback_forward_relative_error": 0.02214777983364559, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.007077484371960163 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438299894332886, + "mean_feedback_forward_cosine": 0.9997444056695507, + "mean_feedback_forward_relative_error": 0.022147701302122687, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0068565161567926405 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438296169042587, + "mean_feedback_forward_cosine": 0.9997444056695507, + "mean_feedback_forward_relative_error": 0.022147619616120093, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.007023111615553498 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382931888103485, + "mean_feedback_forward_cosine": 0.9997444172059337, + "mean_feedback_forward_relative_error": 0.022147543172562314, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006798136008307338 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382905811071396, + "mean_feedback_forward_cosine": 0.9997443999013593, + "mean_feedback_forward_relative_error": 0.02214746941782294, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006447751925587654 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438286855816841, + "mean_feedback_forward_cosine": 0.9997444287423165, + "mean_feedback_forward_relative_error": 0.022147395182400942, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.007190746357142925 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382835030555725, + "mean_feedback_forward_cosine": 0.9997444075922812, + "mean_feedback_forward_relative_error": 0.0221473200757417, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006664027765244246 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382808953523636, + "mean_feedback_forward_cosine": 0.9997444172059337, + "mean_feedback_forward_relative_error": 0.022147241664389448, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006577246445342898 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438278287649155, + "mean_feedback_forward_cosine": 0.9997444191286641, + "mean_feedback_forward_relative_error": 0.022147166377474225, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.007075923771187663 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438274934887886, + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.02214709352401476, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006495891350656748 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382712095975876, + "mean_feedback_forward_cosine": 0.9997444075922812, + "mean_feedback_forward_relative_error": 0.022147022638349764, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006212037367112935 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438268229365349, + "mean_feedback_forward_cosine": 0.9997444018240897, + "mean_feedback_forward_relative_error": 0.022146949514506326, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006336527991443872 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0343826524913311, + "mean_feedback_forward_cosine": 0.9997444133604726, + "mean_feedback_forward_relative_error": 0.02214687269541525, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0067001722053438425 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438260778784752, + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.022146797949267973, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.007536201997920871 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438257798552513, + "mean_feedback_forward_cosine": 0.9997444172059337, + "mean_feedback_forward_relative_error": 0.022146721941328818, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006198162413164973 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382544457912445, + "mean_feedback_forward_cosine": 0.9997444172059337, + "mean_feedback_forward_relative_error": 0.022146650259533236, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005991386559242383 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438250720500946, + "mean_feedback_forward_cosine": 0.9997444152832031, + "mean_feedback_forward_relative_error": 0.022146574777340697, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0070398429016023875 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438247740268707, + "mean_feedback_forward_cosine": 0.9997444095150116, + "mean_feedback_forward_relative_error": 0.02214649931016949, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006825379098318517 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438243642449379, + "mean_feedback_forward_cosine": 0.9997444172059337, + "mean_feedback_forward_relative_error": 0.022146423632699634, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.00649112942956388 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0343824066221714, + "mean_feedback_forward_cosine": 0.9997444287423165, + "mean_feedback_forward_relative_error": 0.022146351635456085, + "min_feedback_forward_cosine": 0.9994107484817505 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.006362385395392776 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382373094558716, + "mean_feedback_forward_cosine": 0.9997444306650469, + "mean_feedback_forward_relative_error": 0.022146280028767163, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.00654777805916965 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03438233584165573, + "mean_feedback_forward_cosine": 0.9997444345105079, + "mean_feedback_forward_relative_error": 0.02214620744569167, + "min_feedback_forward_cosine": 0.99941086769104 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.006772128741443157 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034382302314043045, + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.02214613657504801, + "min_feedback_forward_cosine": 0.9994108080863953 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.006358181383684277 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9223, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.40806366271972655, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2525847040, + "peak_memory_reserved_bytes": 3393191936, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1715624332427979, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10222.548335790634, + "train_wall_s": 10219.140581607819 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684202240000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2745656320000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json new file mode 100644 index 0000000..f4ca482 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json @@ -0,0 +1,3390 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9995934665203094, + "feedback_forward_cosine": [ + 0.9998228549957275, + 0.9998177289962769, + 0.9997299909591675, + 0.9997137784957886, + 0.9997239112854004, + 0.9996533393859863, + 0.9997869729995728, + 0.9996950626373291, + 0.999748706817627, + 0.9996466040611267, + 0.9997207522392273, + 0.9998055696487427, + 0.9997636079788208, + 0.999703049659729, + 0.9997877478599548, + 0.9997124075889587, + 0.9997643232345581, + 0.9996864795684814, + 0.9997842311859131, + 0.9996967315673828, + 0.9997855424880981, + 0.9998668432235718, + 0.9998509883880615, + 0.9998328685760498, + 0.9998459815979004, + 0.9997791647911072, + 0.9998233318328857, + 0.9996689558029175, + 0.9995805621147156, + 0.9992669224739075, + 0.999984860420227 + ], + "feedback_forward_norm_ratio": [ + 0.9996518492698669, + 1.0002232789993286, + 0.9992798566818237, + 1.0017285346984863, + 0.9986801147460938, + 1.0010799169540405, + 0.999582052230835, + 1.0007548332214355, + 0.999423623085022, + 1.0021605491638184, + 0.9986611008644104, + 1.00065016746521, + 0.9993356466293335, + 1.0007761716842651, + 0.999326229095459, + 1.0015517473220825, + 0.9993433952331543, + 1.001268744468689, + 0.9996417760848999, + 1.0017112493515015, + 0.9993442893028259, + 1.0004926919937134, + 0.9998202323913574, + 1.0006990432739258, + 0.9999847412109375, + 1.000831961631775, + 0.9998155236244202, + 1.0010372400283813, + 0.9992583394050598, + 1.0016268491744995, + 1.0027137994766235 + ], + "feedback_forward_relative_error": [ + 0.018819428980350494, + 0.019098827615380287, + 0.023236405104398727, + 0.024010000750422478, + 0.023518972098827362, + 0.026368213817477226, + 0.020639995113015175, + 0.024714823812246323, + 0.022416267544031143, + 0.026700245216488838, + 0.023652229458093643, + 0.019735991954803467, + 0.02174312248826027, + 0.024391446262598038, + 0.020609065890312195, + 0.02405175007879734, + 0.02171194739639759, + 0.02508857101202011, + 0.020772576332092285, + 0.02470984496176243, + 0.020718853920698166, + 0.016328049823641777, + 0.017266981303691864, + 0.01830025017261505, + 0.017553480342030525, + 0.021037161350250244, + 0.018793849274516106, + 0.025765253230929375, + 0.028958963230252266, + 0.03835690766572952, + 0.0061418828554451466 + ], + "innovation_negative_gradient_cosine": [ + 0.9996030330657959, + 0.9995779991149902, + 0.9996092915534973, + 0.9995189905166626, + 0.999632716178894, + 0.9995143413543701, + 0.9996480941772461, + 0.9996263980865479, + 0.9996553659439087, + 0.9995484352111816, + 0.9996688365936279, + 0.9996404051780701, + 0.9996864795684814, + 0.9995951652526855, + 0.9996931552886963, + 0.9996488094329834, + 0.9997095465660095, + 0.9996280074119568, + 0.9997208714485168, + 0.999658465385437, + 0.9997446537017822, + 0.9997918605804443, + 0.999794602394104, + 0.9997910261154175, + 0.9998301267623901, + 0.9998152256011963, + 0.9998809099197388, + 0.9999029636383057, + 0.9998573064804077, + 0.9999117255210876, + 0.8352094292640686 + ], + "mean_feedback_forward_cosine": 0.9997435442862972, + "mean_feedback_forward_relative_error": 0.022103592227663723, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9996030330657959, + 0.9995779991149902, + 0.9996092915534973, + 0.9995189905166626, + 0.999632716178894, + 0.9995143413543701, + 0.9996480941772461, + 0.9996263980865479, + 0.9996553659439087, + 0.9995484352111816, + 0.9996688365936279, + 0.9996404051780701, + 0.9996864795684814, + 0.9995951652526855, + 0.9996931552886963, + 0.9996488094329834, + 0.9997095465660095, + 0.9996280074119568, + 0.9997208714485168, + 0.999658465385437, + 0.9997446537017822, + 0.9997918605804443, + 0.999794602394104, + 0.9997910261154175, + 0.9998301267623901, + 0.9998152256011963, + 0.9998809099197388, + 0.9999029636383057, + 0.9998573064804077, + 0.9999117255210876, + 0.8352094292640686 + ], + "teaching_negative_gradient_cosine": [ + 0.9996030330657959, + 0.9995779991149902, + 0.9996092915534973, + 0.9995189905166626, + 0.999632716178894, + 0.9995143413543701, + 0.9996480941772461, + 0.9996263980865479, + 0.9996553659439087, + 0.9995484352111816, + 0.9996688365936279, + 0.9996404051780701, + 0.9996864795684814, + 0.9995951652526855, + 0.9996931552886963, + 0.9996488094329834, + 0.9997095465660095, + 0.9996280074119568, + 0.9997208714485168, + 0.999658465385437, + 0.9997446537017822, + 0.9997918605804443, + 0.999794602394104, + 0.9997910261154175, + 0.9998301267623901, + 0.9998152256011963, + 0.9998809099197388, + 0.9999029636383057, + 0.9998573064804077, + 0.9999117255210876, + 0.8352094292640686 + ], + "wall_s": 0.10811614990234375 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3494246006011963, + "mean_feedback_forward_cosine": 0.24975507466062422, + "mean_feedback_forward_relative_error": 1.2236636665559584, + "min_feedback_forward_cosine": 0.09657269716262817 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.9218929443359376 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3228424787521362, + "mean_feedback_forward_cosine": 0.33737454731618205, + "mean_feedback_forward_relative_error": 1.1493131826000829, + "min_feedback_forward_cosine": 0.13145390152931213 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4069220595550538 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2915085554122925, + "mean_feedback_forward_cosine": 0.42328307801677334, + "mean_feedback_forward_relative_error": 1.071990292879843, + "min_feedback_forward_cosine": 0.17247258126735687 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.1637197574043274 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.261440396308899, + "mean_feedback_forward_cosine": 0.49132025337988333, + "mean_feedback_forward_relative_error": 1.0071991691666264, + "min_feedback_forward_cosine": 0.2104986310005188 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9531205076599121 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2322450876235962, + "mean_feedback_forward_cosine": 0.5464878394719093, + "mean_feedback_forward_relative_error": 0.9507570064836933, + "min_feedback_forward_cosine": 0.24918709695339203 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8120380458831787 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.209352731704712, + "mean_feedback_forward_cosine": 0.5937906542131978, + "mean_feedback_forward_relative_error": 0.8996120143321252, + "min_feedback_forward_cosine": 0.2835087776184082 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7274218988037109 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1875540018081665, + "mean_feedback_forward_cosine": 0.6319110643479132, + "mean_feedback_forward_relative_error": 0.8559480530600394, + "min_feedback_forward_cosine": 0.31155192852020264 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6619951996421813 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1641260385513306, + "mean_feedback_forward_cosine": 0.6644031443903523, + "mean_feedback_forward_relative_error": 0.8167407426141924, + "min_feedback_forward_cosine": 0.34022173285484314 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6074947142028808 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1407192945480347, + "mean_feedback_forward_cosine": 0.6924694247784153, + "mean_feedback_forward_relative_error": 0.7813257488512224, + "min_feedback_forward_cosine": 0.367876797914505 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.561738102016449 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1186273097991943, + "mean_feedback_forward_cosine": 0.7176954640496161, + "mean_feedback_forward_relative_error": 0.7480449099694529, + "min_feedback_forward_cosine": 0.3937365710735321 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.528632747182846 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0955536365509033, + "mean_feedback_forward_cosine": 0.7401004299040763, + "mean_feedback_forward_relative_error": 0.7170918420437844, + "min_feedback_forward_cosine": 0.4198266863822937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.49865909043312073 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.072723388671875, + "mean_feedback_forward_cosine": 0.7600079726788306, + "mean_feedback_forward_relative_error": 0.6883098050471275, + "min_feedback_forward_cosine": 0.4447500705718994 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4733160824394226 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.04936945438385, + "mean_feedback_forward_cosine": 0.7782817290675256, + "mean_feedback_forward_relative_error": 0.6611209915530297, + "min_feedback_forward_cosine": 0.46964436769485474 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4554216194534302 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0267332792282104, + "mean_feedback_forward_cosine": 0.7951147604373193, + "mean_feedback_forward_relative_error": 0.634850035271337, + "min_feedback_forward_cosine": 0.4923594295978546 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.43665811025619505 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0028632879257202, + "mean_feedback_forward_cosine": 0.8101996471804958, + "mean_feedback_forward_relative_error": 0.6104882330663742, + "min_feedback_forward_cosine": 0.5159362554550171 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.41420850196838377 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.980301558971405, + "mean_feedback_forward_cosine": 0.8237882064234826, + "mean_feedback_forward_relative_error": 0.5876467982607503, + "min_feedback_forward_cosine": 0.5376675128936768 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.396331945734024 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9565905332565308, + "mean_feedback_forward_cosine": 0.8366033742504735, + "mean_feedback_forward_relative_error": 0.5652567576016149, + "min_feedback_forward_cosine": 0.5602827668190002 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3821376348876953 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.933087170124054, + "mean_feedback_forward_cosine": 0.8488402366638184, + "mean_feedback_forward_relative_error": 0.5431818466994071, + "min_feedback_forward_cosine": 0.5823767781257629 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.37093611728668213 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9099282026290894, + "mean_feedback_forward_cosine": 0.8596655476477838, + "mean_feedback_forward_relative_error": 0.5227617837728993, + "min_feedback_forward_cosine": 0.6027913689613342 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3591542248249054 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.887123167514801, + "mean_feedback_forward_cosine": 0.8697121566341769, + "mean_feedback_forward_relative_error": 0.5031659867494337, + "min_feedback_forward_cosine": 0.622233510017395 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3470640205192566 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8634553551673889, + "mean_feedback_forward_cosine": 0.8789754836790024, + "mean_feedback_forward_relative_error": 0.4844576029046889, + "min_feedback_forward_cosine": 0.6421133875846863 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.34042161994934084 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.839503824710846, + "mean_feedback_forward_cosine": 0.8877149147372092, + "mean_feedback_forward_relative_error": 0.4661956309310852, + "min_feedback_forward_cosine": 0.6618831753730774 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3278131375980377 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8167532086372375, + "mean_feedback_forward_cosine": 0.8957876390026461, + "mean_feedback_forward_relative_error": 0.4486217181528768, + "min_feedback_forward_cosine": 0.679898738861084 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.31996051305770873 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7931098341941833, + "mean_feedback_forward_cosine": 0.9032602079453007, + "mean_feedback_forward_relative_error": 0.43187700644616156, + "min_feedback_forward_cosine": 0.698142945766449 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3156246472454071 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7688338756561279, + "mean_feedback_forward_cosine": 0.9103797116587239, + "mean_feedback_forward_relative_error": 0.41535361591846715, + "min_feedback_forward_cosine": 0.7163633108139038 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3047918969631195 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7459856271743774, + "mean_feedback_forward_cosine": 0.9168003509121556, + "mean_feedback_forward_relative_error": 0.39990548020409, + "min_feedback_forward_cosine": 0.7331688404083252 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2932687960529327 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7222238779067993, + "mean_feedback_forward_cosine": 0.9228495224829643, + "mean_feedback_forward_relative_error": 0.38474732637405396, + "min_feedback_forward_cosine": 0.7496653199195862 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2891711379432678 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6987154483795166, + "mean_feedback_forward_cosine": 0.9285203603006178, + "mean_feedback_forward_relative_error": 0.37021790877465277, + "min_feedback_forward_cosine": 0.7655463218688965 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2846348499298096 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6765077114105225, + "mean_feedback_forward_cosine": 0.933629983855832, + "mean_feedback_forward_relative_error": 0.35641476944569617, + "min_feedback_forward_cosine": 0.7801104784011841 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28453598520278933 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6536064743995667, + "mean_feedback_forward_cosine": 0.9384912733108767, + "mean_feedback_forward_relative_error": 0.3428528928948987, + "min_feedback_forward_cosine": 0.7945746183395386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2765980608868599 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.631303608417511, + "mean_feedback_forward_cosine": 0.9429083658802894, + "mean_feedback_forward_relative_error": 0.33018695827453365, + "min_feedback_forward_cosine": 0.808189868927002 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2714606456851959 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.610146164894104, + "mean_feedback_forward_cosine": 0.9470289849465893, + "mean_feedback_forward_relative_error": 0.31787017180073646, + "min_feedback_forward_cosine": 0.8205476999282837 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26684712250709536 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5883633494377136, + "mean_feedback_forward_cosine": 0.9509143098708122, + "mean_feedback_forward_relative_error": 0.30590766403944264, + "min_feedback_forward_cosine": 0.8330889344215393 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2581737526512146 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5682885646820068, + "mean_feedback_forward_cosine": 0.954488661981398, + "mean_feedback_forward_relative_error": 0.2944098338484764, + "min_feedback_forward_cosine": 0.8441348075866699 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2537596423053741 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5482786893844604, + "mean_feedback_forward_cosine": 0.9578218037082303, + "mean_feedback_forward_relative_error": 0.28327399444195533, + "min_feedback_forward_cosine": 0.8548110723495483 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.24909167971611024 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5286539793014526, + "mean_feedback_forward_cosine": 0.9609254348662591, + "mean_feedback_forward_relative_error": 0.2725322518137194, + "min_feedback_forward_cosine": 0.8649052381515503 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2504457321071625 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5099877715110779, + "mean_feedback_forward_cosine": 0.9636899290546295, + "mean_feedback_forward_relative_error": 0.2626277352533033, + "min_feedback_forward_cosine": 0.8742328882217407 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24400218109130858 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4907877445220947, + "mean_feedback_forward_cosine": 0.9664159205652052, + "mean_feedback_forward_relative_error": 0.2525373976076803, + "min_feedback_forward_cosine": 0.8832972645759583 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24301130887031555 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.47337213158607483, + "mean_feedback_forward_cosine": 0.9688678172326857, + "mean_feedback_forward_relative_error": 0.24305977407963045, + "min_feedback_forward_cosine": 0.8913840055465698 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23847964415550232 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4556553363800049, + "mean_feedback_forward_cosine": 0.9711612674497789, + "mean_feedback_forward_relative_error": 0.2338835645106531, + "min_feedback_forward_cosine": 0.8993598222732544 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2351317779827118 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43773984909057617, + "mean_feedback_forward_cosine": 0.9733847475820973, + "mean_feedback_forward_relative_error": 0.2246551436762656, + "min_feedback_forward_cosine": 0.9070427417755127 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23791468878746033 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4209793508052826, + "mean_feedback_forward_cosine": 0.9753380629324144, + "mean_feedback_forward_relative_error": 0.21621643319245307, + "min_feedback_forward_cosine": 0.9139285087585449 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22914538418769836 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40411102771759033, + "mean_feedback_forward_cosine": 0.9771927402865502, + "mean_feedback_forward_relative_error": 0.20793859900966769, + "min_feedback_forward_cosine": 0.9206207394599915 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22515615461349486 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38886141777038574, + "mean_feedback_forward_cosine": 0.9788539909547375, + "mean_feedback_forward_relative_error": 0.20017769115586434, + "min_feedback_forward_cosine": 0.9264025688171387 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22249501668930055 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3738791346549988, + "mean_feedback_forward_cosine": 0.9804486267028316, + "mean_feedback_forward_relative_error": 0.1924592797313967, + "min_feedback_forward_cosine": 0.9318917393684387 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22315158581256866 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3590734899044037, + "mean_feedback_forward_cosine": 0.9819177581417945, + "mean_feedback_forward_relative_error": 0.18506195684594492, + "min_feedback_forward_cosine": 0.9371626973152161 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2209020572423935 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3451373279094696, + "mean_feedback_forward_cosine": 0.9832455381270377, + "mean_feedback_forward_relative_error": 0.1781211506214834, + "min_feedback_forward_cosine": 0.9418581128120422 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21732004253387452 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.331503301858902, + "mean_feedback_forward_cosine": 0.9844934344291687, + "mean_feedback_forward_relative_error": 0.1713354237858326, + "min_feedback_forward_cosine": 0.9463133811950684 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21509123004674913 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3184560537338257, + "mean_feedback_forward_cosine": 0.9856345480488192, + "mean_feedback_forward_relative_error": 0.16490591882217315, + "min_feedback_forward_cosine": 0.9503976106643677 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21304820533752442 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30529871582984924, + "mean_feedback_forward_cosine": 0.9867343152723005, + "mean_feedback_forward_relative_error": 0.15851076764445152, + "min_feedback_forward_cosine": 0.9543863534927368 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21148577075958253 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29255080223083496, + "mean_feedback_forward_cosine": 0.9877838703893846, + "mean_feedback_forward_relative_error": 0.15210549449247698, + "min_feedback_forward_cosine": 0.9580827951431274 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.21210725172042846 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28086167573928833, + "mean_feedback_forward_cosine": 0.9886988747504449, + "mean_feedback_forward_relative_error": 0.1463090545467792, + "min_feedback_forward_cosine": 0.9613544344902039 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2093979245519638 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2697209119796753, + "mean_feedback_forward_cosine": 0.9895473930143541, + "mean_feedback_forward_relative_error": 0.14069656198543887, + "min_feedback_forward_cosine": 0.9643323421478271 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20638913254737853 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2590985596179962, + "mean_feedback_forward_cosine": 0.9903181868214761, + "mean_feedback_forward_relative_error": 0.13540669770971422, + "min_feedback_forward_cosine": 0.9670572280883789 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2027353626680374 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24897614121437073, + "mean_feedback_forward_cosine": 0.9910333944905189, + "mean_feedback_forward_relative_error": 0.13032094749712175, + "min_feedback_forward_cosine": 0.9695420265197754 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1971484166240692 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23913848400115967, + "mean_feedback_forward_cosine": 0.9916987149946151, + "mean_feedback_forward_relative_error": 0.12539815325890818, + "min_feedback_forward_cosine": 0.9718892574310303 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.2015540213871002 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22938434779644012, + "mean_feedback_forward_cosine": 0.9923227448617259, + "mean_feedback_forward_relative_error": 0.12060328981568737, + "min_feedback_forward_cosine": 0.9741147756576538 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20267077679634093 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.220234215259552, + "mean_feedback_forward_cosine": 0.9928988775899333, + "mean_feedback_forward_relative_error": 0.11598796721908354, + "min_feedback_forward_cosine": 0.9761254787445068 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19510222814559935 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21128128468990326, + "mean_feedback_forward_cosine": 0.9934314162500443, + "mean_feedback_forward_relative_error": 0.11157054478122343, + "min_feedback_forward_cosine": 0.9780063629150391 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19726337415695192 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2025374472141266, + "mean_feedback_forward_cosine": 0.9939339449328761, + "mean_feedback_forward_relative_error": 0.1072343870276405, + "min_feedback_forward_cosine": 0.9797813296318054 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1983313562297821 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19448445737361908, + "mean_feedback_forward_cosine": 0.9943843233969903, + "mean_feedback_forward_relative_error": 0.10317202777631822, + "min_feedback_forward_cosine": 0.9813475608825684 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1898783901500702 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18635331094264984, + "mean_feedback_forward_cosine": 0.9948251804997844, + "mean_feedback_forward_relative_error": 0.09904818537254487, + "min_feedback_forward_cosine": 0.9828630685806274 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19620646570205688 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1787845939397812, + "mean_feedback_forward_cosine": 0.9952079942149501, + "mean_feedback_forward_relative_error": 0.09532134523314814, + "min_feedback_forward_cosine": 0.984218418598175 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.19244548971176148 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17157258093357086, + "mean_feedback_forward_cosine": 0.995572809250124, + "mean_feedback_forward_relative_error": 0.09162182679339763, + "min_feedback_forward_cosine": 0.9854565858840942 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18765464908599855 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1646691858768463, + "mean_feedback_forward_cosine": 0.9959113328687607, + "mean_feedback_forward_relative_error": 0.08806144157724996, + "min_feedback_forward_cosine": 0.9865995645523071 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18908571057319642 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15825730562210083, + "mean_feedback_forward_cosine": 0.9962189274449502, + "mean_feedback_forward_relative_error": 0.08468100735016408, + "min_feedback_forward_cosine": 0.9876154661178589 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18638300099372865 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15199866890907288, + "mean_feedback_forward_cosine": 0.996499728771948, + "mean_feedback_forward_relative_error": 0.08147101512839718, + "min_feedback_forward_cosine": 0.9885689616203308 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1860867762851715 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1457824558019638, + "mean_feedback_forward_cosine": 0.9967687264565499, + "mean_feedback_forward_relative_error": 0.07829492887662302, + "min_feedback_forward_cosine": 0.9894801378250122 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18519178348064422 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1397344470024109, + "mean_feedback_forward_cosine": 0.9970131401092776, + "mean_feedback_forward_relative_error": 0.07528625374599811, + "min_feedback_forward_cosine": 0.990331768989563 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18635893783569335 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13424205780029297, + "mean_feedback_forward_cosine": 0.997237301641895, + "mean_feedback_forward_relative_error": 0.07240463306586589, + "min_feedback_forward_cosine": 0.9910707473754883 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.183787780418396 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12852050364017487, + "mean_feedback_forward_cosine": 0.9974454006841106, + "mean_feedback_forward_relative_error": 0.06964071479535872, + "min_feedback_forward_cosine": 0.9918137788772583 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1831598755478859 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12334217876195908, + "mean_feedback_forward_cosine": 0.9976358356014374, + "mean_feedback_forward_relative_error": 0.06700158954387711, + "min_feedback_forward_cosine": 0.9924593567848206 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1804471286201477 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11835367977619171, + "mean_feedback_forward_cosine": 0.9978146033902322, + "mean_feedback_forward_relative_error": 0.06443115107474788, + "min_feedback_forward_cosine": 0.9930556416511536 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17463825380325318 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11329466849565506, + "mean_feedback_forward_cosine": 0.9979841824500791, + "mean_feedback_forward_relative_error": 0.06190058997561855, + "min_feedback_forward_cosine": 0.9936355352401733 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.18227414899349212 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10864241421222687, + "mean_feedback_forward_cosine": 0.998138023960975, + "mean_feedback_forward_relative_error": 0.059497022821057226, + "min_feedback_forward_cosine": 0.9941459894180298 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17801999889135361 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10435745120048523, + "mean_feedback_forward_cosine": 0.9982780121987865, + "mean_feedback_forward_relative_error": 0.05721727512296169, + "min_feedback_forward_cosine": 0.9945970773696899 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1765562902069092 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10019416362047195, + "mean_feedback_forward_cosine": 0.9984062794716128, + "mean_feedback_forward_relative_error": 0.05505009404113216, + "min_feedback_forward_cosine": 0.9950186014175415 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17533604710578918 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09605756402015686, + "mean_feedback_forward_cosine": 0.9985280709881936, + "mean_feedback_forward_relative_error": 0.05291419536356003, + "min_feedback_forward_cosine": 0.9954196810722351 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17509580305576325 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09225407242774963, + "mean_feedback_forward_cosine": 0.9986377935255727, + "mean_feedback_forward_relative_error": 0.05090813531029609, + "min_feedback_forward_cosine": 0.9957756996154785 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17200049627304076 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08852459490299225, + "mean_feedback_forward_cosine": 0.9987405211694779, + "mean_feedback_forward_relative_error": 0.04895767881985633, + "min_feedback_forward_cosine": 0.9961096048355103 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1740963316631317 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0847882628440857, + "mean_feedback_forward_cosine": 0.9988403108812147, + "mean_feedback_forward_relative_error": 0.0469820047578504, + "min_feedback_forward_cosine": 0.9964291453361511 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17462056126117706 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.081540547311306, + "mean_feedback_forward_cosine": 0.9989246245353453, + "mean_feedback_forward_relative_error": 0.04524247964183169, + "min_feedback_forward_cosine": 0.9966965317726135 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16816279454231262 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07826635986566544, + "mean_feedback_forward_cosine": 0.9990054138245121, + "mean_feedback_forward_relative_error": 0.043515207546372565, + "min_feedback_forward_cosine": 0.9969558715820312 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16755252022266387 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07517033070325851, + "mean_feedback_forward_cosine": 0.9990779449862819, + "mean_feedback_forward_relative_error": 0.0419047734730186, + "min_feedback_forward_cosine": 0.9971915483474731 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16846428961753845 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07203090190887451, + "mean_feedback_forward_cosine": 0.9991485553403054, + "mean_feedback_forward_relative_error": 0.04027359788456271, + "min_feedback_forward_cosine": 0.9974207282066345 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.1688399866914749 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06910812854766846, + "mean_feedback_forward_cosine": 0.999212589956099, + "mean_feedback_forward_relative_error": 0.03873308156166346, + "min_feedback_forward_cosine": 0.9976263046264648 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16943029206752777 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06635339558124542, + "mean_feedback_forward_cosine": 0.9992720696233934, + "mean_feedback_forward_relative_error": 0.0372469199280585, + "min_feedback_forward_cosine": 0.9978106021881104 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1678766533613205 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06360498815774918, + "mean_feedback_forward_cosine": 0.9993283556353662, + "mean_feedback_forward_relative_error": 0.035780609765600775, + "min_feedback_forward_cosine": 0.9979878664016724 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16858385267257692 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06101856008172035, + "mean_feedback_forward_cosine": 0.9993801578398673, + "mean_feedback_forward_relative_error": 0.03437901651787181, + "min_feedback_forward_cosine": 0.9981483221054077 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16932012516021727 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05856332555413246, + "mean_feedback_forward_cosine": 0.9994261822392864, + "mean_feedback_forward_relative_error": 0.03308020258742955, + "min_feedback_forward_cosine": 0.9982941746711731 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16621017497062682 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056274961680173874, + "mean_feedback_forward_cosine": 0.9994692244837361, + "mean_feedback_forward_relative_error": 0.031815457938899915, + "min_feedback_forward_cosine": 0.9984242916107178 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16870774178981782 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05400526523590088, + "mean_feedback_forward_cosine": 0.9995097979422538, + "mean_feedback_forward_relative_error": 0.03057598082288619, + "min_feedback_forward_cosine": 0.9985485672950745 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16830255025863647 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05168907344341278, + "mean_feedback_forward_cosine": 0.9995479910604416, + "mean_feedback_forward_relative_error": 0.02936676538158809, + "min_feedback_forward_cosine": 0.998670220375061 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.17085209632396697 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04972684755921364, + "mean_feedback_forward_cosine": 0.9995804256008517, + "mean_feedback_forward_relative_error": 0.028296017298294653, + "min_feedback_forward_cosine": 0.9987689256668091 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1595303696537018 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04777975007891655, + "mean_feedback_forward_cosine": 0.9996111931339386, + "mean_feedback_forward_relative_error": 0.027240780242268118, + "min_feedback_forward_cosine": 0.9988634586334229 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1591718906545639 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04582907259464264, + "mean_feedback_forward_cosine": 0.9996409127789159, + "mean_feedback_forward_relative_error": 0.026182132983399976, + "min_feedback_forward_cosine": 0.9989540576934814 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.15775668437480928 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.043918970972299576, + "mean_feedback_forward_cosine": 0.9996690192530232, + "mean_feedback_forward_relative_error": 0.025140733878699043, + "min_feedback_forward_cosine": 0.9990392923355103 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.16501512162208556 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.042085129767656326, + "mean_feedback_forward_cosine": 0.9996941474176222, + "mean_feedback_forward_relative_error": 0.02417049887439897, + "min_feedback_forward_cosine": 0.9991177320480347 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1663829922580719 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.040321219712495804, + "mean_feedback_forward_cosine": 0.9997173393926313, + "mean_feedback_forward_relative_error": 0.023241238458262335, + "min_feedback_forward_cosine": 0.9991900324821472 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16541705914497376 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038703419268131256, + "mean_feedback_forward_cosine": 0.9997384528959951, + "mean_feedback_forward_relative_error": 0.022359017237660386, + "min_feedback_forward_cosine": 0.9992535710334778 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1572692034816742 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03869009017944336, + "mean_feedback_forward_cosine": 0.9997386201735465, + "mean_feedback_forward_relative_error": 0.022349788747247187, + "min_feedback_forward_cosine": 0.9992542266845703 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09018189778804779 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03868260979652405, + "mean_feedback_forward_cosine": 0.9997387355373751, + "mean_feedback_forward_relative_error": 0.02234361457428144, + "min_feedback_forward_cosine": 0.9992544651031494 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.0702357754611969 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038675710558891296, + "mean_feedback_forward_cosine": 0.9997388489784733, + "mean_feedback_forward_relative_error": 0.02233777842634628, + "min_feedback_forward_cosine": 0.9992546439170837 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05911954815030098 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038668934255838394, + "mean_feedback_forward_cosine": 0.9997389431922666, + "mean_feedback_forward_relative_error": 0.02233229835908259, + "min_feedback_forward_cosine": 0.9992549419403076 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.053982924233675005 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03866267576813698, + "mean_feedback_forward_cosine": 0.9997390393287905, + "mean_feedback_forward_relative_error": 0.022327132688294494, + "min_feedback_forward_cosine": 0.9992551803588867 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04768824928998947 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038656026124954224, + "mean_feedback_forward_cosine": 0.9997391373880448, + "mean_feedback_forward_relative_error": 0.022322015085768317, + "min_feedback_forward_cosine": 0.9992554187774658 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.044762988986968996 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03864984214305878, + "mean_feedback_forward_cosine": 0.9997392354472991, + "mean_feedback_forward_relative_error": 0.022317053509816047, + "min_feedback_forward_cosine": 0.9992556571960449 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.04090550207406282 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03864375129342079, + "mean_feedback_forward_cosine": 0.9997393162019791, + "mean_feedback_forward_relative_error": 0.022312285679001963, + "min_feedback_forward_cosine": 0.9992560148239136 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03672991137981415 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03863723203539848, + "mean_feedback_forward_cosine": 0.9997393873430067, + "mean_feedback_forward_relative_error": 0.022307376302177866, + "min_feedback_forward_cosine": 0.9992561340332031 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03636346148490906 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03863116726279259, + "mean_feedback_forward_cosine": 0.999739489247722, + "mean_feedback_forward_relative_error": 0.022302621720178474, + "min_feedback_forward_cosine": 0.9992564916610718 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03307025036215782 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0386250838637352, + "mean_feedback_forward_cosine": 0.9997395642342106, + "mean_feedback_forward_relative_error": 0.02229790753054042, + "min_feedback_forward_cosine": 0.9992567300796509 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.032560025746524335 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03861873596906662, + "mean_feedback_forward_cosine": 0.9997396680616564, + "mean_feedback_forward_relative_error": 0.022293061754035372, + "min_feedback_forward_cosine": 0.9992568492889404 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.031159462106227876 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038612477481365204, + "mean_feedback_forward_cosine": 0.9997397622754497, + "mean_feedback_forward_relative_error": 0.02228844101210275, + "min_feedback_forward_cosine": 0.9992570877075195 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.02895025592982769 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03860635310411453, + "mean_feedback_forward_cosine": 0.9997398526437821, + "mean_feedback_forward_relative_error": 0.022283784038717707, + "min_feedback_forward_cosine": 0.9992573857307434 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.027717424998283387 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038599804043769836, + "mean_feedback_forward_cosine": 0.9997399487803059, + "mean_feedback_forward_relative_error": 0.022278990461340836, + "min_feedback_forward_cosine": 0.9992576837539673 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02742766058623791 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0385938324034214, + "mean_feedback_forward_cosine": 0.9997400372259079, + "mean_feedback_forward_relative_error": 0.022274345550085266, + "min_feedback_forward_cosine": 0.9992578029632568 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.02480702959358692 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03858798369765282, + "mean_feedback_forward_cosine": 0.9997401179805878, + "mean_feedback_forward_relative_error": 0.022269865693224054, + "min_feedback_forward_cosine": 0.9992581605911255 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.024096279681921004 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03858125954866409, + "mean_feedback_forward_cosine": 0.9997402064261898, + "mean_feedback_forward_relative_error": 0.0222650226506014, + "min_feedback_forward_cosine": 0.9992583990097046 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02406107524037361 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03857496753334999, + "mean_feedback_forward_cosine": 0.999740319867288, + "mean_feedback_forward_relative_error": 0.02226024840567862, + "min_feedback_forward_cosine": 0.9992586374282837 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.022692273498773573 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03856874629855156, + "mean_feedback_forward_cosine": 0.9997404025446984, + "mean_feedback_forward_relative_error": 0.022255517061679594, + "min_feedback_forward_cosine": 0.9992587566375732 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.02152857553243637 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03856196999549866, + "mean_feedback_forward_cosine": 0.9997404948357613, + "mean_feedback_forward_relative_error": 0.022250843399594867, + "min_feedback_forward_cosine": 0.9992590546607971 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.021097033315598965 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03855599835515022, + "mean_feedback_forward_cosine": 0.9997405813586328, + "mean_feedback_forward_relative_error": 0.022246450426117066, + "min_feedback_forward_cosine": 0.999259352684021 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.019415753941237925 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038549765944480896, + "mean_feedback_forward_cosine": 0.999740679417887, + "mean_feedback_forward_relative_error": 0.022241957680953127, + "min_feedback_forward_cosine": 0.9992594718933105 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018513377546072005 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038543179631233215, + "mean_feedback_forward_cosine": 0.9997407717089499, + "mean_feedback_forward_relative_error": 0.022237125033092116, + "min_feedback_forward_cosine": 0.9992598295211792 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.019798150712773206 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03853682428598404, + "mean_feedback_forward_cosine": 0.9997408832273176, + "mean_feedback_forward_relative_error": 0.022232338395570556, + "min_feedback_forward_cosine": 0.9992600679397583 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.019051756429374216 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03853025287389755, + "mean_feedback_forward_cosine": 0.9997409524456147, + "mean_feedback_forward_relative_error": 0.022227739990358392, + "min_feedback_forward_cosine": 0.9992603063583374 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01761170121088624 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03852350637316704, + "mean_feedback_forward_cosine": 0.9997410678094433, + "mean_feedback_forward_relative_error": 0.0222228912696723, + "min_feedback_forward_cosine": 0.999260425567627 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.018012608473673464 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03851630166172981, + "mean_feedback_forward_cosine": 0.9997411639459671, + "mean_feedback_forward_relative_error": 0.02221776469941101, + "min_feedback_forward_cosine": 0.9992607831954956 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.018129425482228397 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03850993141531944, + "mean_feedback_forward_cosine": 0.9997412773870653, + "mean_feedback_forward_relative_error": 0.022213073372240027, + "min_feedback_forward_cosine": 0.9992610216140747 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.016865740612745286 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03850335255265236, + "mean_feedback_forward_cosine": 0.9997413504508234, + "mean_feedback_forward_relative_error": 0.022208275228378274, + "min_feedback_forward_cosine": 0.9992613196372986 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.016044229831546544 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03849764168262482, + "mean_feedback_forward_cosine": 0.9997414331282338, + "mean_feedback_forward_relative_error": 0.022204190267310027, + "min_feedback_forward_cosine": 0.9992614984512329 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.014237593904286623 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03849118947982788, + "mean_feedback_forward_cosine": 0.9997415408011405, + "mean_feedback_forward_relative_error": 0.022199535321804784, + "min_feedback_forward_cosine": 0.999261736869812 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.015662225911915302 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03848356753587723, + "mean_feedback_forward_cosine": 0.9997416446285863, + "mean_feedback_forward_relative_error": 0.022194101505221858, + "min_feedback_forward_cosine": 0.9992619752883911 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.015363968058675528 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038476429879665375, + "mean_feedback_forward_cosine": 0.9997417638378758, + "mean_feedback_forward_relative_error": 0.022188984173079653, + "min_feedback_forward_cosine": 0.9992623329162598 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.015120945150852204 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0384698286652565, + "mean_feedback_forward_cosine": 0.9997418618971302, + "mean_feedback_forward_relative_error": 0.022184129383775494, + "min_feedback_forward_cosine": 0.9992625713348389 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013894460175037385 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038462795317173004, + "mean_feedback_forward_cosine": 0.9997419811064198, + "mean_feedback_forward_relative_error": 0.022179397814456495, + "min_feedback_forward_cosine": 0.999262809753418 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.014200187674462796 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03845585510134697, + "mean_feedback_forward_cosine": 0.9997420753202131, + "mean_feedback_forward_relative_error": 0.022174607391559308, + "min_feedback_forward_cosine": 0.9992630481719971 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.014059381691217423 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03844912350177765, + "mean_feedback_forward_cosine": 0.9997421599203541, + "mean_feedback_forward_relative_error": 0.022169834213151086, + "min_feedback_forward_cosine": 0.999263346195221 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.013133898133039475 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038440823554992676, + "mean_feedback_forward_cosine": 0.9997422964342179, + "mean_feedback_forward_relative_error": 0.022164240374319976, + "min_feedback_forward_cosine": 0.9992636442184448 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.015220713586062192 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03843340277671814, + "mean_feedback_forward_cosine": 0.9997424233344293, + "mean_feedback_forward_relative_error": 0.022158963921209498, + "min_feedback_forward_cosine": 0.9992640018463135 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.013820327893048524 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038426365703344345, + "mean_feedback_forward_cosine": 0.9997424983209179, + "mean_feedback_forward_relative_error": 0.02215411449452081, + "min_feedback_forward_cosine": 0.9992642998695374 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.012681063927710056 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03841947391629219, + "mean_feedback_forward_cosine": 0.9997426232983989, + "mean_feedback_forward_relative_error": 0.02214897439003952, + "min_feedback_forward_cosine": 0.9992644786834717 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.012453560272157192 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038412805646657944, + "mean_feedback_forward_cosine": 0.9997427309713056, + "mean_feedback_forward_relative_error": 0.02214435871029573, + "min_feedback_forward_cosine": 0.9992647171020508 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011589504393190146 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03840521723031998, + "mean_feedback_forward_cosine": 0.9997428347987514, + "mean_feedback_forward_relative_error": 0.02213895676897899, + "min_feedback_forward_cosine": 0.9992650747299194 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.012354984072893859 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0383986234664917, + "mean_feedback_forward_cosine": 0.9997429443943885, + "mean_feedback_forward_relative_error": 0.02213434368792561, + "min_feedback_forward_cosine": 0.9992653727531433 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010935525473169982 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03839060664176941, + "mean_feedback_forward_cosine": 0.9997430482218342, + "mean_feedback_forward_relative_error": 0.022128906491543015, + "min_feedback_forward_cosine": 0.9992656707763672 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.012370056481957435 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03838308900594711, + "mean_feedback_forward_cosine": 0.9997431751220457, + "mean_feedback_forward_relative_error": 0.022123666600354256, + "min_feedback_forward_cosine": 0.9992658495903015 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.011595427350848913 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038375068455934525, + "mean_feedback_forward_cosine": 0.9997432770267609, + "mean_feedback_forward_relative_error": 0.02211842517698965, + "min_feedback_forward_cosine": 0.9992662668228149 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.011534525494948029 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03836674615740776, + "mean_feedback_forward_cosine": 0.99974339431332, + "mean_feedback_forward_relative_error": 0.022112570748093627, + "min_feedback_forward_cosine": 0.999266505241394 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.012372618612647057 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835928067564964, + "mean_feedback_forward_cosine": 0.9997435462090277, + "mean_feedback_forward_relative_error": 0.022107558430082375, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.0101877372437343 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835919499397278, + "mean_feedback_forward_cosine": 0.9997435173680705, + "mean_feedback_forward_relative_error": 0.022107452905226137, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.01016308972582221 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038359135389328, + "mean_feedback_forward_cosine": 0.9997435077544181, + "mean_feedback_forward_relative_error": 0.022107359772968675, + "min_feedback_forward_cosine": 0.9992667436599731 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.009141140254288912 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835907578468323, + "mean_feedback_forward_cosine": 0.9997435000634962, + "mean_feedback_forward_relative_error": 0.02210726635530591, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008568081556633115 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038359012454748154, + "mean_feedback_forward_cosine": 0.9997435077544181, + "mean_feedback_forward_relative_error": 0.022107177429021366, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.00920205535236746 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358960300683975, + "mean_feedback_forward_cosine": 0.9997435327499143, + "mean_feedback_forward_relative_error": 0.022107091852493824, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.008157383770346641 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358911871910095, + "mean_feedback_forward_cosine": 0.9997435135226096, + "mean_feedback_forward_relative_error": 0.022107006786691565, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007641669589877128 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358863443136215, + "mean_feedback_forward_cosine": 0.9997435404408362, + "mean_feedback_forward_relative_error": 0.022106922456934568, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007855133966505527 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358818739652634, + "mean_feedback_forward_cosine": 0.9997435250589924, + "mean_feedback_forward_relative_error": 0.022106839824588068, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.007312930462509394 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358770310878754, + "mean_feedback_forward_cosine": 0.99974351544534, + "mean_feedback_forward_relative_error": 0.02210676003127329, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007122721250653267 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358718156814575, + "mean_feedback_forward_cosine": 0.999743526981723, + "mean_feedback_forward_relative_error": 0.022106672652185924, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.008205709009766578 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0383586660027504, + "mean_feedback_forward_cosine": 0.9997435250589924, + "mean_feedback_forward_relative_error": 0.022106589524135474, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007589992175102234 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835861757397652, + "mean_feedback_forward_cosine": 0.9997435212135315, + "mean_feedback_forward_relative_error": 0.022106507507663582, + "min_feedback_forward_cosine": 0.9992667436599731 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007849129783064127 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835856169462204, + "mean_feedback_forward_cosine": 0.999743526981723, + "mean_feedback_forward_relative_error": 0.022106419572786938, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007677926018461585 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835850954055786, + "mean_feedback_forward_cosine": 0.9997435212135315, + "mean_feedback_forward_relative_error": 0.022106339013384234, + "min_feedback_forward_cosine": 0.9992667436599731 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007710860101804137 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835846483707428, + "mean_feedback_forward_cosine": 0.999743526981723, + "mean_feedback_forward_relative_error": 0.022106256290909745, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.00716620623677969 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0383584089577198, + "mean_feedback_forward_cosine": 0.9997435462090277, + "mean_feedback_forward_relative_error": 0.022106170413955566, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.00753851116836071 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835836052894592, + "mean_feedback_forward_cosine": 0.9997435346726449, + "mean_feedback_forward_relative_error": 0.022106088337398345, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007372459765523672 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038358304649591446, + "mean_feedback_forward_cosine": 0.9997435423635668, + "mean_feedback_forward_relative_error": 0.022106010902432666, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006897408326864243 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835826367139816, + "mean_feedback_forward_cosine": 0.9997435289044534, + "mean_feedback_forward_relative_error": 0.02210593456402421, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006711386793404817 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835821896791458, + "mean_feedback_forward_cosine": 0.9997435442862972, + "mean_feedback_forward_relative_error": 0.022105860493836865, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006610966650322079 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0383581705391407, + "mean_feedback_forward_cosine": 0.9997435385181058, + "mean_feedback_forward_relative_error": 0.02210578015975414, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.007155023907870054 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835812211036682, + "mean_feedback_forward_cosine": 0.9997435308271839, + "mean_feedback_forward_relative_error": 0.02210570598441747, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006892853389084339 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835806995630264, + "mean_feedback_forward_cosine": 0.9997435462090277, + "mean_feedback_forward_relative_error": 0.022105625575228084, + "min_feedback_forward_cosine": 0.9992669820785522 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0068606297035515304 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835802152752876, + "mean_feedback_forward_cosine": 0.9997435423635668, + "mean_feedback_forward_relative_error": 0.022105549372011615, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006844044959545136 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835797309875488, + "mean_feedback_forward_cosine": 0.9997435250589924, + "mean_feedback_forward_relative_error": 0.022105472192408576, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0070141338916122915 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0383579358458519, + "mean_feedback_forward_cosine": 0.9997435365953753, + "mean_feedback_forward_relative_error": 0.022105397746687936, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006231199810951949 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357898592948914, + "mean_feedback_forward_cosine": 0.9997435385181058, + "mean_feedback_forward_relative_error": 0.02210531798341582, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.007026433295011521 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835785388946533, + "mean_feedback_forward_cosine": 0.9997435404408362, + "mean_feedback_forward_relative_error": 0.022105244168591116, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0062881049451977014 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835780918598175, + "mean_feedback_forward_cosine": 0.9997435346726449, + "mean_feedback_forward_relative_error": 0.02210516375940173, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006777262418195605 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835776820778847, + "mean_feedback_forward_cosine": 0.99974355582268, + "mean_feedback_forward_relative_error": 0.022105090335131652, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0069646171709895136 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835771977901459, + "mean_feedback_forward_cosine": 0.9997435423635668, + "mean_feedback_forward_relative_error": 0.022105010481731545, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006164614319801331 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357678800821304, + "mean_feedback_forward_cosine": 0.9997435481317581, + "mean_feedback_forward_relative_error": 0.02210493797376271, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006379115580543875 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835761547088623, + "mean_feedback_forward_cosine": 0.9997435519772191, + "mean_feedback_forward_relative_error": 0.022104859502325134, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006984152169078588 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835757076740265, + "mean_feedback_forward_cosine": 0.9997435519772191, + "mean_feedback_forward_relative_error": 0.022104779618882364, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0067542794423550365 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835752233862877, + "mean_feedback_forward_cosine": 0.9997435308271839, + "mean_feedback_forward_relative_error": 0.022104701553020748, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006420430699884892 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357485085725784, + "mean_feedback_forward_cosine": 0.9997435404408362, + "mean_feedback_forward_relative_error": 0.022104620993618044, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0062022028070688245 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357436656951904, + "mean_feedback_forward_cosine": 0.9997435538999496, + "mean_feedback_forward_relative_error": 0.022104545676660155, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006654117411822081 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357388228178024, + "mean_feedback_forward_cosine": 0.9997435500544887, + "mean_feedback_forward_relative_error": 0.022104472703030034, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006000056919679046 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357336074113846, + "mean_feedback_forward_cosine": 0.9997435423635668, + "mean_feedback_forward_relative_error": 0.022104391677966042, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.007019033769890666 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835730627179146, + "mean_feedback_forward_cosine": 0.9997435519772191, + "mean_feedback_forward_relative_error": 0.0221043243072927, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006082673249095678 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835725784301758, + "mean_feedback_forward_cosine": 0.9997435692817934, + "mean_feedback_forward_relative_error": 0.02210424972638007, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006554380866661668 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835722431540489, + "mean_feedback_forward_cosine": 0.9997435635136019, + "mean_feedback_forward_relative_error": 0.022104179381483024, + "min_feedback_forward_cosine": 0.9992669820785522 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005760983594581485 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835718333721161, + "mean_feedback_forward_cosine": 0.9997435385181058, + "mean_feedback_forward_relative_error": 0.022104102427199963, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006248740261644125 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357146084308624, + "mean_feedback_forward_cosine": 0.9997435308271839, + "mean_feedback_forward_relative_error": 0.022104029844124472, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.00558479504339397 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835710138082504, + "mean_feedback_forward_cosine": 0.9997435481317581, + "mean_feedback_forward_relative_error": 0.02210396020523002, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.00593255756765604 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835706040263176, + "mean_feedback_forward_cosine": 0.9997435712045238, + "mean_feedback_forward_relative_error": 0.022103888974074396, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005618409241810441 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038357026875019073, + "mean_feedback_forward_cosine": 0.9997435769727153, + "mean_feedback_forward_relative_error": 0.022103817893132087, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.005956166023015976 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038356974720954895, + "mean_feedback_forward_cosine": 0.9997435500544887, + "mean_feedback_forward_relative_error": 0.022103737018281413, + "min_feedback_forward_cosine": 0.9992668032646179 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.006610512352585793 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835694119334221, + "mean_feedback_forward_cosine": 0.9997435423635668, + "mean_feedback_forward_relative_error": 0.022103664119757952, + "min_feedback_forward_cosine": 0.9992668628692627 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005906309532299638 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03835690766572952, + "mean_feedback_forward_cosine": 0.9997435442862972, + "mean_feedback_forward_relative_error": 0.022103592227663723, + "min_feedback_forward_cosine": 0.9992669224739075 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005890884000062943 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9242, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3821306213378906, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2525847040, + "peak_memory_reserved_bytes": 3393191936, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1649303436279297, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10230.376488685608, + "train_wall_s": 10227.093840837479 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684202240000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2745656320000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json new file mode 100644 index 0000000..4b02437 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json @@ -0,0 +1,3390 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9996226906776429, + "feedback_forward_cosine": [ + 0.9998250007629395, + 0.9997978210449219, + 0.999758780002594, + 0.9997544884681702, + 0.9997519254684448, + 0.9997063875198364, + 0.9997261166572571, + 0.9996814131736755, + 0.9998026490211487, + 0.9997053146362305, + 0.9997532367706299, + 0.9998248815536499, + 0.9997637867927551, + 0.9997279047966003, + 0.9997736215591431, + 0.9996997117996216, + 0.9997929334640503, + 0.999712347984314, + 0.9997790455818176, + 0.9996895790100098, + 0.999791145324707, + 0.9998652935028076, + 0.9998343586921692, + 0.9998188018798828, + 0.9998387694358826, + 0.9997784495353699, + 0.9998390674591064, + 0.9996950626373291, + 0.9995816946029663, + 0.9992325305938721, + 0.9999851584434509 + ], + "feedback_forward_norm_ratio": [ + 0.9995194673538208, + 1.0007712841033936, + 1.000235676765442, + 1.0012747049331665, + 0.9986703991889954, + 1.0010285377502441, + 0.9991890788078308, + 1.0012438297271729, + 0.999777615070343, + 1.0017437934875488, + 0.9986539483070374, + 1.0003631114959717, + 0.9993255734443665, + 1.0008310079574585, + 0.999375581741333, + 1.0016775131225586, + 0.9995991587638855, + 1.0015108585357666, + 0.9994063973426819, + 1.0014119148254395, + 0.9992547035217285, + 1.0004510879516602, + 1.0000050067901611, + 1.0006206035614014, + 0.9998761415481567, + 1.0008827447891235, + 0.9999377727508545, + 1.0009171962738037, + 0.9990943074226379, + 1.0017184019088745, + 1.0025044679641724 + ], + "feedback_forward_relative_error": [ + 0.018711449578404427, + 0.0201310645788908, + 0.021970022469758987, + 0.02221059985458851, + 0.022298455238342285, + 0.024271324276924133, + 0.023405304178595543, + 0.025290008634328842, + 0.019864164292812347, + 0.0243570227175951, + 0.022242913022637367, + 0.018726056441664696, + 0.02173955924808979, + 0.023355519399046898, + 0.021280093118548393, + 0.02458626963198185, + 0.020351605489850044, + 0.02405170165002346, + 0.02102426066994667, + 0.0249736737459898, + 0.020445581525564194, + 0.01641915924847126, + 0.018208028748631477, + 0.019054405391216278, + 0.01795344054698944, + 0.021074611693620682, + 0.017935434356331825, + 0.024722829461097717, + 0.028928402811288834, + 0.039251793175935745, + 0.005994729697704315 + ], + "innovation_negative_gradient_cosine": [ + 0.9996582269668579, + 0.9996131062507629, + 0.9996418356895447, + 0.9995585680007935, + 0.9996452331542969, + 0.9995571374893188, + 0.9996507167816162, + 0.9995861053466797, + 0.9996627569198608, + 0.9996532201766968, + 0.9996786713600159, + 0.9996798038482666, + 0.99970543384552, + 0.9996029138565063, + 0.9997167587280273, + 0.9996203780174255, + 0.9997353553771973, + 0.9996602535247803, + 0.9997515082359314, + 0.9996718168258667, + 0.9997636675834656, + 0.9998162388801575, + 0.9998011589050293, + 0.9997760057449341, + 0.999826192855835, + 0.9998064041137695, + 0.999868631362915, + 0.9999010562896729, + 0.9998774528503418, + 0.9999389052391052, + 0.7473083734512329 + ], + "mean_feedback_forward_cosine": 0.9997512025217856, + "mean_feedback_forward_relative_error": 0.021768693061124895, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9996582269668579, + 0.9996131062507629, + 0.9996418356895447, + 0.9995585680007935, + 0.9996452331542969, + 0.9995571374893188, + 0.9996507167816162, + 0.9995861053466797, + 0.9996627569198608, + 0.9996532201766968, + 0.9996786713600159, + 0.9996798038482666, + 0.99970543384552, + 0.9996029138565063, + 0.9997167587280273, + 0.9996203780174255, + 0.9997353553771973, + 0.9996602535247803, + 0.9997515082359314, + 0.9996718168258667, + 0.9997636675834656, + 0.9998162388801575, + 0.9998011589050293, + 0.9997760057449341, + 0.999826192855835, + 0.9998064041137695, + 0.999868631362915, + 0.9999010562896729, + 0.9998774528503418, + 0.9999389052391052, + 0.7473083734512329 + ], + "teaching_negative_gradient_cosine": [ + 0.9996582269668579, + 0.9996131062507629, + 0.9996418356895447, + 0.9995585680007935, + 0.9996452331542969, + 0.9995571374893188, + 0.9996507167816162, + 0.9995861053466797, + 0.9996627569198608, + 0.9996532201766968, + 0.9996786713600159, + 0.9996798038482666, + 0.99970543384552, + 0.9996029138565063, + 0.9997167587280273, + 0.9996203780174255, + 0.9997353553771973, + 0.9996602535247803, + 0.9997515082359314, + 0.9996718168258667, + 0.9997636675834656, + 0.9998162388801575, + 0.9998011589050293, + 0.9997760057449341, + 0.999826192855835, + 0.9998064041137695, + 0.999868631362915, + 0.9999010562896729, + 0.9998774528503418, + 0.9999389052391052, + 0.7473083734512329 + ], + "wall_s": 0.11745548248291016 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3349723815917969, + "mean_feedback_forward_cosine": 0.25990806856463033, + "mean_feedback_forward_relative_error": 1.2173136838020817, + "min_feedback_forward_cosine": 0.11319214105606079 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.884043900756836 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3094651699066162, + "mean_feedback_forward_cosine": 0.35645093456391364, + "mean_feedback_forward_relative_error": 1.135014666664985, + "min_feedback_forward_cosine": 0.14877766370773315 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.2980965474700927 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2844831943511963, + "mean_feedback_forward_cosine": 0.43406235979449365, + "mean_feedback_forward_relative_error": 1.0651125821375078, + "min_feedback_forward_cosine": 0.18230189383029938 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0457662846374511 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.257002353668213, + "mean_feedback_forward_cosine": 0.49849749909293267, + "mean_feedback_forward_relative_error": 1.003033523598025, + "min_feedback_forward_cosine": 0.21811392903327942 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8763960188293457 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2295236587524414, + "mean_feedback_forward_cosine": 0.5516888105100201, + "mean_feedback_forward_relative_error": 0.9484026335900829, + "min_feedback_forward_cosine": 0.2541167736053467 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7647238268280029 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2094680070877075, + "mean_feedback_forward_cosine": 0.5949192950802464, + "mean_feedback_forward_relative_error": 0.9011446881678796, + "min_feedback_forward_cosine": 0.2851111888885498 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6871170258522034 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1871272325515747, + "mean_feedback_forward_cosine": 0.6318055352857036, + "mean_feedback_forward_relative_error": 0.8588615530921567, + "min_feedback_forward_cosine": 0.31310462951660156 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6291039484786988 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1641530990600586, + "mean_feedback_forward_cosine": 0.6638145408322734, + "mean_feedback_forward_relative_error": 0.8201376699632213, + "min_feedback_forward_cosine": 0.34058821201324463 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.586492281703949 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1416432857513428, + "mean_feedback_forward_cosine": 0.692069120945469, + "mean_feedback_forward_relative_error": 0.7841187965485358, + "min_feedback_forward_cosine": 0.366452693939209 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5476818246841431 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1194217205047607, + "mean_feedback_forward_cosine": 0.7173972206730996, + "mean_feedback_forward_relative_error": 0.7505404612710399, + "min_feedback_forward_cosine": 0.39185771346092224 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5080024182033539 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.09666907787323, + "mean_feedback_forward_cosine": 0.7399982973452537, + "mean_feedback_forward_relative_error": 0.7192039864678537, + "min_feedback_forward_cosine": 0.4173521399497986 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.48509688959121705 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0738338232040405, + "mean_feedback_forward_cosine": 0.7605622458842493, + "mean_feedback_forward_relative_error": 0.6895902368330187, + "min_feedback_forward_cosine": 0.44172707200050354 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4655131332588196 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0519495010375977, + "mean_feedback_forward_cosine": 0.7789299305408232, + "mean_feedback_forward_relative_error": 0.6620327253495494, + "min_feedback_forward_cosine": 0.4650299847126007 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.43883312452316287 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.029091238975525, + "mean_feedback_forward_cosine": 0.7959011793136597, + "mean_feedback_forward_relative_error": 0.6354356696528773, + "min_feedback_forward_cosine": 0.488097608089447 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.419735015001297 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0058889389038086, + "mean_feedback_forward_cosine": 0.8113211816357028, + "mean_feedback_forward_relative_error": 0.6103085418862682, + "min_feedback_forward_cosine": 0.5115365386009216 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4059742552089691 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9821122288703918, + "mean_feedback_forward_cosine": 0.8253101264276812, + "mean_feedback_forward_relative_error": 0.5864743201001998, + "min_feedback_forward_cosine": 0.5349057912826538 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3919783997154236 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9576853513717651, + "mean_feedback_forward_cosine": 0.8383109781049913, + "mean_feedback_forward_relative_error": 0.5636168324178265, + "min_feedback_forward_cosine": 0.5582489967346191 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.379794213848114 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9333812594413757, + "mean_feedback_forward_cosine": 0.85005771921527, + "mean_feedback_forward_relative_error": 0.5421345978975296, + "min_feedback_forward_cosine": 0.5805637836456299 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.36320647946357726 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9090774059295654, + "mean_feedback_forward_cosine": 0.8611171495529913, + "mean_feedback_forward_relative_error": 0.5212958705040717, + "min_feedback_forward_cosine": 0.6023303270339966 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.353560187292099 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8852086067199707, + "mean_feedback_forward_cosine": 0.8712598642995281, + "mean_feedback_forward_relative_error": 0.5012798193962343, + "min_feedback_forward_cosine": 0.6232945322990417 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.34309122184753416 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8593584895133972, + "mean_feedback_forward_cosine": 0.88098979573096, + "mean_feedback_forward_relative_error": 0.4814876466989517, + "min_feedback_forward_cosine": 0.6450381278991699 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.33722845947265623 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8355581760406494, + "mean_feedback_forward_cosine": 0.8898247191982884, + "mean_feedback_forward_relative_error": 0.462748872656976, + "min_feedback_forward_cosine": 0.6645044684410095 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.33217840114593505 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8125510811805725, + "mean_feedback_forward_cosine": 0.8976805344704659, + "mean_feedback_forward_relative_error": 0.4454753009542342, + "min_feedback_forward_cosine": 0.6828471422195435 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3130986285972595 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7896296977996826, + "mean_feedback_forward_cosine": 0.9050561285788014, + "mean_feedback_forward_relative_error": 0.42864807623047985, + "min_feedback_forward_cosine": 0.700555682182312 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.309137520236969 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7661307454109192, + "mean_feedback_forward_cosine": 0.9120974483028534, + "mean_feedback_forward_relative_error": 0.41205001982950396, + "min_feedback_forward_cosine": 0.7182934880256653 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3054877910041809 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7443239688873291, + "mean_feedback_forward_cosine": 0.918309542440599, + "mean_feedback_forward_relative_error": 0.3968265316178722, + "min_feedback_forward_cosine": 0.7340960502624512 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.29228196764945985 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7201147079467773, + "mean_feedback_forward_cosine": 0.9244374152152769, + "mean_feedback_forward_relative_error": 0.3812758355371414, + "min_feedback_forward_cosine": 0.7511111497879028 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29109219769001005 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6977641582489014, + "mean_feedback_forward_cosine": 0.9298505129352692, + "mean_feedback_forward_relative_error": 0.3670203181043748, + "min_feedback_forward_cosine": 0.7661524415016174 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2771931728935242 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.67481929063797, + "mean_feedback_forward_cosine": 0.9349553546597881, + "mean_feedback_forward_relative_error": 0.3530991913810853, + "min_feedback_forward_cosine": 0.7812139391899109 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.279468224697113 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6530283689498901, + "mean_feedback_forward_cosine": 0.9397015263957362, + "mean_feedback_forward_relative_error": 0.3397381053336205, + "min_feedback_forward_cosine": 0.7948744893074036 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2755542581796646 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6322877407073975, + "mean_feedback_forward_cosine": 0.9440194733681218, + "mean_feedback_forward_relative_error": 0.3271398010753816, + "min_feedback_forward_cosine": 0.8076740503311157 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2665172424221039 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6114274263381958, + "mean_feedback_forward_cosine": 0.9480447096209372, + "mean_feedback_forward_relative_error": 0.31496808125126746, + "min_feedback_forward_cosine": 0.8199342489242554 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2590750113105774 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5898762941360474, + "mean_feedback_forward_cosine": 0.9518554499072414, + "mean_feedback_forward_relative_error": 0.3030531540993721, + "min_feedback_forward_cosine": 0.8323711156845093 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.25794359429359437 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5693039894104004, + "mean_feedback_forward_cosine": 0.9554501272016956, + "mean_feedback_forward_relative_error": 0.2913327791517781, + "min_feedback_forward_cosine": 0.8437407612800598 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2553165550708771 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.548820972442627, + "mean_feedback_forward_cosine": 0.9587425454970329, + "mean_feedback_forward_relative_error": 0.2802302210081008, + "min_feedback_forward_cosine": 0.8547742366790771 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2507095138835907 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5298911333084106, + "mean_feedback_forward_cosine": 0.9617095435819318, + "mean_feedback_forward_relative_error": 0.26985916903903406, + "min_feedback_forward_cosine": 0.8644689321517944 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24415148473739623 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5102511048316956, + "mean_feedback_forward_cosine": 0.9645711414275631, + "mean_feedback_forward_relative_error": 0.25947804532704816, + "min_feedback_forward_cosine": 0.8741466999053955 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24547552498340608 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4911539554595947, + "mean_feedback_forward_cosine": 0.9672493127084547, + "mean_feedback_forward_relative_error": 0.24944197963322362, + "min_feedback_forward_cosine": 0.8833778500556946 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2402279775428772 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4727519154548645, + "mean_feedback_forward_cosine": 0.9697137782650609, + "mean_feedback_forward_relative_error": 0.2398086798287207, + "min_feedback_forward_cosine": 0.891808271408081 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23981824064254761 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45590081810951233, + "mean_feedback_forward_cosine": 0.9719306768909577, + "mean_feedback_forward_relative_error": 0.23079736578849055, + "min_feedback_forward_cosine": 0.8993138074874878 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22953084101200102 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4394187033176422, + "mean_feedback_forward_cosine": 0.9739993022334191, + "mean_feedback_forward_relative_error": 0.22205424044401414, + "min_feedback_forward_cosine": 0.9063918590545654 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2279604758644104 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42284128069877625, + "mean_feedback_forward_cosine": 0.975942728980895, + "mean_feedback_forward_relative_error": 0.21353178952009447, + "min_feedback_forward_cosine": 0.9132277965545654 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22551626369476319 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4067952036857605, + "mean_feedback_forward_cosine": 0.9777397224980016, + "mean_feedback_forward_relative_error": 0.20538089159996278, + "min_feedback_forward_cosine": 0.9196112751960754 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22367396668434142 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3910422623157501, + "mean_feedback_forward_cosine": 0.9794375992590382, + "mean_feedback_forward_relative_error": 0.19732192158699036, + "min_feedback_forward_cosine": 0.9256578683853149 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22419112510681152 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37633296847343445, + "mean_feedback_forward_cosine": 0.9809703115494021, + "mean_feedback_forward_relative_error": 0.18982052658834764, + "min_feedback_forward_cosine": 0.9310845732688904 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22211388847827912 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3624928593635559, + "mean_feedback_forward_cosine": 0.9823756525593419, + "mean_feedback_forward_relative_error": 0.18263642261585883, + "min_feedback_forward_cosine": 0.9360270500183105 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21705526567935943 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3488244116306305, + "mean_feedback_forward_cosine": 0.9836807289431172, + "mean_feedback_forward_relative_error": 0.1757090908865775, + "min_feedback_forward_cosine": 0.9407151937484741 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2158580997800827 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33559462428092957, + "mean_feedback_forward_cosine": 0.9848859194786318, + "mean_feedback_forward_relative_error": 0.16906738485540113, + "min_feedback_forward_cosine": 0.9450581073760986 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20908656046390534 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3229137659072876, + "mean_feedback_forward_cosine": 0.985992964236967, + "mean_feedback_forward_relative_error": 0.1627447511399946, + "min_feedback_forward_cosine": 0.9490985870361328 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2055481154346466 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31016919016838074, + "mean_feedback_forward_cosine": 0.9870364396802841, + "mean_feedback_forward_relative_error": 0.1565605353924536, + "min_feedback_forward_cosine": 0.9530023336410522 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20949283747196198 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29767873883247375, + "mean_feedback_forward_cosine": 0.9880552368779336, + "mean_feedback_forward_relative_error": 0.15028610104514706, + "min_feedback_forward_cosine": 0.9566665887832642 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2089853828907013 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28635820746421814, + "mean_feedback_forward_cosine": 0.988928867924598, + "mean_feedback_forward_relative_error": 0.14466954539379767, + "min_feedback_forward_cosine": 0.9598673582077026 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2060586162662506 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2751348316669464, + "mean_feedback_forward_cosine": 0.9897554882111088, + "mean_feedback_forward_relative_error": 0.1391651259074288, + "min_feedback_forward_cosine": 0.9629354476928711 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20208517848968505 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2640094459056854, + "mean_feedback_forward_cosine": 0.9905401083730883, + "mean_feedback_forward_relative_error": 0.13374620291494554, + "min_feedback_forward_cosine": 0.9658530950546265 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2008401836013794 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25394725799560547, + "mean_feedback_forward_cosine": 0.9912350619992902, + "mean_feedback_forward_relative_error": 0.12873109333938168, + "min_feedback_forward_cosine": 0.9683619141578674 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1994219180393219 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24407565593719482, + "mean_feedback_forward_cosine": 0.9918945912391909, + "mean_feedback_forward_relative_error": 0.12379781685529216, + "min_feedback_forward_cosine": 0.9707534909248352 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19782180733680724 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23445096611976624, + "mean_feedback_forward_cosine": 0.9924968769473415, + "mean_feedback_forward_relative_error": 0.11911300689943376, + "min_feedback_forward_cosine": 0.9729940295219421 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19733688462734222 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22489391267299652, + "mean_feedback_forward_cosine": 0.9930661935960093, + "mean_feedback_forward_relative_error": 0.11452286486183444, + "min_feedback_forward_cosine": 0.9751298427581787 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19768200149536133 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2162090688943863, + "mean_feedback_forward_cosine": 0.9935784993633148, + "mean_feedback_forward_relative_error": 0.11020198860956777, + "min_feedback_forward_cosine": 0.9769930243492126 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.1947573218727112 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20737668871879578, + "mean_feedback_forward_cosine": 0.9940720200538635, + "mean_feedback_forward_relative_error": 0.10589810833334923, + "min_feedback_forward_cosine": 0.9788262844085693 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1977950550365448 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19919180870056152, + "mean_feedback_forward_cosine": 0.9945168860497013, + "mean_feedback_forward_relative_error": 0.1018484249230354, + "min_feedback_forward_cosine": 0.980451226234436 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18924341267585754 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19078901410102844, + "mean_feedback_forward_cosine": 0.9949399925047352, + "mean_feedback_forward_relative_error": 0.09785510803903302, + "min_feedback_forward_cosine": 0.9820586442947388 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19221665453910827 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18314522504806519, + "mean_feedback_forward_cosine": 0.995325994107031, + "mean_feedback_forward_relative_error": 0.09406068644696666, + "min_feedback_forward_cosine": 0.9834667444229126 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18745509509086608 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17564021050930023, + "mean_feedback_forward_cosine": 0.995681478131202, + "mean_feedback_forward_relative_error": 0.09041920724895693, + "min_feedback_forward_cosine": 0.9847860932350159 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1897912615942955 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16852696239948273, + "mean_feedback_forward_cosine": 0.9960067464459327, + "mean_feedback_forward_relative_error": 0.08695372224094407, + "min_feedback_forward_cosine": 0.9859851598739624 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1850361007261276 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16218271851539612, + "mean_feedback_forward_cosine": 0.9962990860785207, + "mean_feedback_forward_relative_error": 0.08371402034836431, + "min_feedback_forward_cosine": 0.987011194229126 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18250540663719178 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15562225878238678, + "mean_feedback_forward_cosine": 0.9965787145399279, + "mean_feedback_forward_relative_error": 0.08049640516119619, + "min_feedback_forward_cosine": 0.9880364537239075 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18282743438720703 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14947938919067383, + "mean_feedback_forward_cosine": 0.9968364181057099, + "mean_feedback_forward_relative_error": 0.07741088065649232, + "min_feedback_forward_cosine": 0.9889557361602783 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.183723773355484 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1434447169303894, + "mean_feedback_forward_cosine": 0.9970758730365384, + "mean_feedback_forward_relative_error": 0.07443589410714564, + "min_feedback_forward_cosine": 0.9898234009742737 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18212517969369887 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13758878409862518, + "mean_feedback_forward_cosine": 0.9972983521799887, + "mean_feedback_forward_relative_error": 0.07155920655256318, + "min_feedback_forward_cosine": 0.9906343221664429 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.18086355746269225 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13201770186424255, + "mean_feedback_forward_cosine": 0.9975032037304293, + "mean_feedback_forward_relative_error": 0.06879527920917157, + "min_feedback_forward_cosine": 0.9913747310638428 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17983569902420043 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12645262479782104, + "mean_feedback_forward_cosine": 0.997694036652965, + "mean_feedback_forward_relative_error": 0.0661307120275113, + "min_feedback_forward_cosine": 0.9920851588249207 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.18159530221939088 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12123321741819382, + "mean_feedback_forward_cosine": 0.99786796492915, + "mean_feedback_forward_relative_error": 0.06359651883042627, + "min_feedback_forward_cosine": 0.9927232265472412 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1807903523159027 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11612481623888016, + "mean_feedback_forward_cosine": 0.9980316719701213, + "mean_feedback_forward_relative_error": 0.06111879724889032, + "min_feedback_forward_cosine": 0.9933205842971802 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.18251712452888488 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11155027896165848, + "mean_feedback_forward_cosine": 0.9981785474285003, + "mean_feedback_forward_relative_error": 0.058794992585336006, + "min_feedback_forward_cosine": 0.9938347339630127 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17368552847385407 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10704890638589859, + "mean_feedback_forward_cosine": 0.9983175185418898, + "mean_feedback_forward_relative_error": 0.05651775657409622, + "min_feedback_forward_cosine": 0.994320273399353 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17487866772651672 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10279550403356552, + "mean_feedback_forward_cosine": 0.998444099580088, + "mean_feedback_forward_relative_error": 0.054354885953568646, + "min_feedback_forward_cosine": 0.9947606325149536 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1732005853843689 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09852416068315506, + "mean_feedback_forward_cosine": 0.9985643740623228, + "mean_feedback_forward_relative_error": 0.05222089306241082, + "min_feedback_forward_cosine": 0.9951850771903992 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17545873102188111 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09449601918458939, + "mean_feedback_forward_cosine": 0.9986736409125789, + "mean_feedback_forward_relative_error": 0.050204635207210815, + "min_feedback_forward_cosine": 0.995569109916687 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17256650600910187 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0906398594379425, + "mean_feedback_forward_cosine": 0.9987740881981388, + "mean_feedback_forward_relative_error": 0.04827549933425842, + "min_feedback_forward_cosine": 0.9959231615066528 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17154016873359682 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08689554035663605, + "mean_feedback_forward_cosine": 0.9988671041304066, + "mean_feedback_forward_relative_error": 0.04641639303055502, + "min_feedback_forward_cosine": 0.9962524771690369 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16824539291858673 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08307069540023804, + "mean_feedback_forward_cosine": 0.9989579704500013, + "mean_feedback_forward_relative_error": 0.04452944841356047, + "min_feedback_forward_cosine": 0.9965732097625732 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1754700458049774 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0797470211982727, + "mean_feedback_forward_cosine": 0.9990356256884914, + "mean_feedback_forward_relative_error": 0.04284472327919737, + "min_feedback_forward_cosine": 0.996841311454773 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1721455662059784 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07657480239868164, + "mean_feedback_forward_cosine": 0.9991070608938893, + "mean_feedback_forward_relative_error": 0.04123145305821972, + "min_feedback_forward_cosine": 0.9970871210098267 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16954058514118195 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07354815304279327, + "mean_feedback_forward_cosine": 0.9991730297765424, + "mean_feedback_forward_relative_error": 0.039683275405437715, + "min_feedback_forward_cosine": 0.9973117709159851 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.1681050411939621 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07066227495670319, + "mean_feedback_forward_cosine": 0.9992358280766395, + "mean_feedback_forward_relative_error": 0.038148833378668756, + "min_feedback_forward_cosine": 0.9975184202194214 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16526709414482116 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06776793301105499, + "mean_feedback_forward_cosine": 0.9992965325232475, + "mean_feedback_forward_relative_error": 0.03660518351581789, + "min_feedback_forward_cosine": 0.9977167844772339 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.17173692792892456 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06512687355279922, + "mean_feedback_forward_cosine": 0.9993474541171905, + "mean_feedback_forward_relative_error": 0.03525468307517229, + "min_feedback_forward_cosine": 0.9978905916213989 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.1647601191496849 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06248985603451729, + "mean_feedback_forward_cosine": 0.9993976912190837, + "mean_feedback_forward_relative_error": 0.03387635940265271, + "min_feedback_forward_cosine": 0.9980573058128357 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16422447180747987 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05999685078859329, + "mean_feedback_forward_cosine": 0.999443123417516, + "mean_feedback_forward_relative_error": 0.03257730176612254, + "min_feedback_forward_cosine": 0.9982086420059204 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16520101567268372 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.057540878653526306, + "mean_feedback_forward_cosine": 0.9994849216553473, + "mean_feedback_forward_relative_error": 0.031336988262351484, + "min_feedback_forward_cosine": 0.9983521699905396 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16500281002998352 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05531076341867447, + "mean_feedback_forward_cosine": 0.9995224148996414, + "mean_feedback_forward_relative_error": 0.0301763474821083, + "min_feedback_forward_cosine": 0.9984773397445679 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1594796088385582 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.053001850843429565, + "mean_feedback_forward_cosine": 0.9995596620344347, + "mean_feedback_forward_relative_error": 0.02898126714412243, + "min_feedback_forward_cosine": 0.9986019730567932 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16494166981697084 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05076294764876366, + "mean_feedback_forward_cosine": 0.9995936539865309, + "mean_feedback_forward_relative_error": 0.02784653329440663, + "min_feedback_forward_cosine": 0.9987173080444336 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16599833611011505 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048671893775463104, + "mean_feedback_forward_cosine": 0.9996238696959711, + "mean_feedback_forward_relative_error": 0.026796938941603708, + "min_feedback_forward_cosine": 0.9988206028938293 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16241411400318145 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.046662647277116776, + "mean_feedback_forward_cosine": 0.999652999062692, + "mean_feedback_forward_relative_error": 0.025740397973887382, + "min_feedback_forward_cosine": 0.9989158511161804 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16649711584091187 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04476156458258629, + "mean_feedback_forward_cosine": 0.9996800115031581, + "mean_feedback_forward_relative_error": 0.024722059707968466, + "min_feedback_forward_cosine": 0.9990021586418152 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1602955338382721 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04304143413901329, + "mean_feedback_forward_cosine": 0.9997028458502984, + "mean_feedback_forward_relative_error": 0.023824551503264135, + "min_feedback_forward_cosine": 0.9990774989128113 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15782464220046996 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04127826541662216, + "mean_feedback_forward_cosine": 0.9997257974839979, + "mean_feedback_forward_relative_error": 0.022887508444968733, + "min_feedback_forward_cosine": 0.9991511702537537 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16260396190643311 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03956778347492218, + "mean_feedback_forward_cosine": 0.9997466398823646, + "mean_feedback_forward_relative_error": 0.022003036307831927, + "min_feedback_forward_cosine": 0.9992200136184692 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1612073497080803 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03955405205488205, + "mean_feedback_forward_cosine": 0.9997468052371856, + "mean_feedback_forward_relative_error": 0.021993932133961107, + "min_feedback_forward_cosine": 0.9992206692695618 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08828540871143341 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039547037333250046, + "mean_feedback_forward_cosine": 0.9997469148328227, + "mean_feedback_forward_relative_error": 0.021988024349294363, + "min_feedback_forward_cosine": 0.9992209672927856 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06392889021635055 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039540424942970276, + "mean_feedback_forward_cosine": 0.9997469936647723, + "mean_feedback_forward_relative_error": 0.02198252376289137, + "min_feedback_forward_cosine": 0.9992212057113647 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05590131332099438 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03953411057591438, + "mean_feedback_forward_cosine": 0.9997471071058704, + "mean_feedback_forward_relative_error": 0.02197735247412516, + "min_feedback_forward_cosine": 0.9992215037345886 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04892089093506336 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03952792286872864, + "mean_feedback_forward_cosine": 0.9997472128560466, + "mean_feedback_forward_relative_error": 0.0219724063701447, + "min_feedback_forward_cosine": 0.9992217421531677 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.043798088161945345 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03952197730541229, + "mean_feedback_forward_cosine": 0.9997472705379609, + "mean_feedback_forward_relative_error": 0.021967468467811422, + "min_feedback_forward_cosine": 0.999221920967102 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04162949063181877 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039516009390354156, + "mean_feedback_forward_cosine": 0.99974734744718, + "mean_feedback_forward_relative_error": 0.021962743012174483, + "min_feedback_forward_cosine": 0.9992221593856812 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03833890787601471 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03951011225581169, + "mean_feedback_forward_cosine": 0.9997474416609733, + "mean_feedback_forward_relative_error": 0.02195797023934222, + "min_feedback_forward_cosine": 0.9992223978042603 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03592932426214218 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03950371965765953, + "mean_feedback_forward_cosine": 0.9997475224156533, + "mean_feedback_forward_relative_error": 0.021953170142707326, + "min_feedback_forward_cosine": 0.9992227554321289 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.035366734300255775 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0394979789853096, + "mean_feedback_forward_cosine": 0.9997476012476029, + "mean_feedback_forward_relative_error": 0.021948661625144944, + "min_feedback_forward_cosine": 0.9992228746414185 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.030628358398377895 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039492469280958176, + "mean_feedback_forward_cosine": 0.9997476935386658, + "mean_feedback_forward_relative_error": 0.02194429343686469, + "min_feedback_forward_cosine": 0.9992231130599976 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.029325532891750335 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03948645293712616, + "mean_feedback_forward_cosine": 0.9997477839069981, + "mean_feedback_forward_relative_error": 0.021939773322834123, + "min_feedback_forward_cosine": 0.9992232918739319 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.029357269192934035 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03948098048567772, + "mean_feedback_forward_cosine": 0.9997478608162172, + "mean_feedback_forward_relative_error": 0.021935472595355204, + "min_feedback_forward_cosine": 0.9992235898971558 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.026094600776433944 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03947504609823227, + "mean_feedback_forward_cosine": 0.9997479415708973, + "mean_feedback_forward_relative_error": 0.021931041062118545, + "min_feedback_forward_cosine": 0.9992238283157349 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02651489243924618 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039468780159950256, + "mean_feedback_forward_cosine": 0.9997480242483078, + "mean_feedback_forward_relative_error": 0.02192651267133413, + "min_feedback_forward_cosine": 0.999224066734314 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02624331370830536 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039462897926568985, + "mean_feedback_forward_cosine": 0.9997481069257182, + "mean_feedback_forward_relative_error": 0.021922080432094874, + "min_feedback_forward_cosine": 0.9992242455482483 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024278797640800476 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039457205682992935, + "mean_feedback_forward_cosine": 0.9997481857576678, + "mean_feedback_forward_relative_error": 0.02191784540792146, + "min_feedback_forward_cosine": 0.9992244243621826 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.022229577784240246 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03945150226354599, + "mean_feedback_forward_cosine": 0.9997482799714611, + "mean_feedback_forward_relative_error": 0.021913597675701304, + "min_feedback_forward_cosine": 0.9992246031761169 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02159910560131073 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0394456572830677, + "mean_feedback_forward_cosine": 0.9997483549579498, + "mean_feedback_forward_relative_error": 0.02190916364892356, + "min_feedback_forward_cosine": 0.999224841594696 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.021546049241423608 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03943952918052673, + "mean_feedback_forward_cosine": 0.9997484472490126, + "mean_feedback_forward_relative_error": 0.02190469577908516, + "min_feedback_forward_cosine": 0.9992251396179199 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.021970328575372695 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03943442925810814, + "mean_feedback_forward_cosine": 0.9997485318491536, + "mean_feedback_forward_relative_error": 0.021900775301600655, + "min_feedback_forward_cosine": 0.999225378036499 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.017999292779564857 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039428383111953735, + "mean_feedback_forward_cosine": 0.9997486126038336, + "mean_feedback_forward_relative_error": 0.021896474138503114, + "min_feedback_forward_cosine": 0.9992256164550781 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.018960947802364826 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0394226498901844, + "mean_feedback_forward_cosine": 0.9997487202767403, + "mean_feedback_forward_relative_error": 0.0218921925632223, + "min_feedback_forward_cosine": 0.999225914478302 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018468987340182066 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039417196065187454, + "mean_feedback_forward_cosine": 0.9997487798813851, + "mean_feedback_forward_relative_error": 0.02188799973397005, + "min_feedback_forward_cosine": 0.9992260932922363 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.016994322579205037 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039411451667547226, + "mean_feedback_forward_cosine": 0.9997488490996822, + "mean_feedback_forward_relative_error": 0.02188394600225072, + "min_feedback_forward_cosine": 0.9992262721061707 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.017177284666597844 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039405569434165955, + "mean_feedback_forward_cosine": 0.999748949081667, + "mean_feedback_forward_relative_error": 0.021879751670865283, + "min_feedback_forward_cosine": 0.999226450920105 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01637926060795784 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03939993306994438, + "mean_feedback_forward_cosine": 0.9997490086863118, + "mean_feedback_forward_relative_error": 0.02187567081062063, + "min_feedback_forward_cosine": 0.9992266893386841 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.015048764585107566 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039394695311784744, + "mean_feedback_forward_cosine": 0.9997491086682966, + "mean_feedback_forward_relative_error": 0.021871750483349446, + "min_feedback_forward_cosine": 0.9992269277572632 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.014844148950874806 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03938871994614601, + "mean_feedback_forward_cosine": 0.9997491605820195, + "mean_feedback_forward_relative_error": 0.02186735495624523, + "min_feedback_forward_cosine": 0.9992271661758423 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.015637938914895056 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03938230872154236, + "mean_feedback_forward_cosine": 0.9997492721003871, + "mean_feedback_forward_relative_error": 0.021862772293388844, + "min_feedback_forward_cosine": 0.9992274045944214 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.015976422900632024 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03937559574842453, + "mean_feedback_forward_cosine": 0.9997494028460595, + "mean_feedback_forward_relative_error": 0.021858133405687347, + "min_feedback_forward_cosine": 0.9992275834083557 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.015033096481263638 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03936991095542908, + "mean_feedback_forward_cosine": 0.9997494566825128, + "mean_feedback_forward_relative_error": 0.021853979556791244, + "min_feedback_forward_cosine": 0.9992278218269348 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.013741299457252026 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03936389833688736, + "mean_feedback_forward_cosine": 0.9997495451281148, + "mean_feedback_forward_relative_error": 0.021849748873782736, + "min_feedback_forward_cosine": 0.9992280602455139 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013263638440668583 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03935781866312027, + "mean_feedback_forward_cosine": 0.9997496181918729, + "mean_feedback_forward_relative_error": 0.02184542185897308, + "min_feedback_forward_cosine": 0.9992283582687378 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.013716067549288273 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03935127705335617, + "mean_feedback_forward_cosine": 0.9997497297102406, + "mean_feedback_forward_relative_error": 0.02184095559641719, + "min_feedback_forward_cosine": 0.9992285966873169 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.01336184063643217 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03934511914849281, + "mean_feedback_forward_cosine": 0.9997498085421901, + "mean_feedback_forward_relative_error": 0.021836617841355262, + "min_feedback_forward_cosine": 0.9992287158966064 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.013175583146512508 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039338357746601105, + "mean_feedback_forward_cosine": 0.99974988929687, + "mean_feedback_forward_relative_error": 0.021831967462334904, + "min_feedback_forward_cosine": 0.9992291331291199 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.01325496376425028 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03933264687657356, + "mean_feedback_forward_cosine": 0.9997499758197416, + "mean_feedback_forward_relative_error": 0.02182770337188436, + "min_feedback_forward_cosine": 0.999229371547699 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.011746058116555214 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03932664915919304, + "mean_feedback_forward_cosine": 0.9997500834926483, + "mean_feedback_forward_relative_error": 0.021823468017241648, + "min_feedback_forward_cosine": 0.9992295503616333 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.011617636746764182 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03932031989097595, + "mean_feedback_forward_cosine": 0.9997501873200939, + "mean_feedback_forward_relative_error": 0.021819020591435895, + "min_feedback_forward_cosine": 0.9992297887802124 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01183700292006135 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03931359946727753, + "mean_feedback_forward_cosine": 0.9997502776884264, + "mean_feedback_forward_relative_error": 0.02181434832633503, + "min_feedback_forward_cosine": 0.9992299675941467 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.012418612503856421 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03930603340268135, + "mean_feedback_forward_cosine": 0.9997503795931416, + "mean_feedback_forward_relative_error": 0.021809443335739837, + "min_feedback_forward_cosine": 0.9992303848266602 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.01267631758466363 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03929959982633591, + "mean_feedback_forward_cosine": 0.999750466116013, + "mean_feedback_forward_relative_error": 0.021804695378147786, + "min_feedback_forward_cosine": 0.9992306232452393 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.01170698377624154 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03929335996508598, + "mean_feedback_forward_cosine": 0.999750550716154, + "mean_feedback_forward_relative_error": 0.02180038312930734, + "min_feedback_forward_cosine": 0.9992308020591736 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.010822395001798869 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03928709775209427, + "mean_feedback_forward_cosine": 0.9997506680027131, + "mean_feedback_forward_relative_error": 0.021795969892052874, + "min_feedback_forward_cosine": 0.9992311000823975 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010739987428039312 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03928085416555405, + "mean_feedback_forward_cosine": 0.9997507718301588, + "mean_feedback_forward_relative_error": 0.021791406035903963, + "min_feedback_forward_cosine": 0.9992313385009766 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010964248820766806 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03927396982908249, + "mean_feedback_forward_cosine": 0.9997508621984913, + "mean_feedback_forward_relative_error": 0.02178660711093295, + "min_feedback_forward_cosine": 0.9992315769195557 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.011047148571312428 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03926721587777138, + "mean_feedback_forward_cosine": 0.999750962180476, + "mean_feedback_forward_relative_error": 0.02178180754004467, + "min_feedback_forward_cosine": 0.9992318749427795 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.01048615313872695 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03926093131303787, + "mean_feedback_forward_cosine": 0.9997510583169998, + "mean_feedback_forward_relative_error": 0.021777256256750515, + "min_feedback_forward_cosine": 0.9992320537567139 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.009832702446728944 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253976196050644, + "mean_feedback_forward_cosine": 0.9997511563762542, + "mean_feedback_forward_relative_error": 0.021772383486912135, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.010426850525736808 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925387188792229, + "mean_feedback_forward_cosine": 0.9997511506080627, + "mean_feedback_forward_relative_error": 0.021772278577930505, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.009929457244426012 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925381973385811, + "mean_feedback_forward_cosine": 0.9997511506080627, + "mean_feedback_forward_relative_error": 0.021772196531415947, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008251808605119586 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925376012921333, + "mean_feedback_forward_cosine": 0.9997511525307933, + "mean_feedback_forward_relative_error": 0.021772107770366055, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008543098836168647 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925371170043945, + "mean_feedback_forward_cosine": 0.9997511582989846, + "mean_feedback_forward_relative_error": 0.02177202533329687, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.00813495438784361 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253655821084976, + "mean_feedback_forward_cosine": 0.9997511544535237, + "mean_feedback_forward_relative_error": 0.021771943076483665, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.007969489766955376 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925361484289169, + "mean_feedback_forward_cosine": 0.9997511525307933, + "mean_feedback_forward_relative_error": 0.02177186980242691, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007356664264649153 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253562688827515, + "mean_feedback_forward_cosine": 0.9997511486853322, + "mean_feedback_forward_relative_error": 0.021771795101343625, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007573647451475263 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253514260053635, + "mean_feedback_forward_cosine": 0.9997511640671761, + "mean_feedback_forward_relative_error": 0.021771720129876368, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.007154074359536171 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253465831279755, + "mean_feedback_forward_cosine": 0.9997511602217152, + "mean_feedback_forward_relative_error": 0.021771639885921634, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0077800188639760015 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925343230366707, + "mean_feedback_forward_cosine": 0.9997511698353675, + "mean_feedback_forward_relative_error": 0.021771564493857084, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.006432479541748762 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925338014960289, + "mean_feedback_forward_cosine": 0.999751171758098, + "mean_feedback_forward_relative_error": 0.021771485857184854, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007193801200687885 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253346621990204, + "mean_feedback_forward_cosine": 0.9997511602217152, + "mean_feedback_forward_relative_error": 0.02177141136640022, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.006947582246512174 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925330564379692, + "mean_feedback_forward_cosine": 0.999751171758098, + "mean_feedback_forward_relative_error": 0.021771336409954295, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.006525301996283233 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925326094031334, + "mean_feedback_forward_cosine": 0.9997511390716799, + "mean_feedback_forward_relative_error": 0.021771262655214917, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007157000288069248 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925321251153946, + "mean_feedback_forward_cosine": 0.9997511659899065, + "mean_feedback_forward_relative_error": 0.021771182366196188, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.00680718580968678 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253171533346176, + "mean_feedback_forward_cosine": 0.9997511563762542, + "mean_feedback_forward_relative_error": 0.021771103459139986, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.007068686625491828 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039253126829862595, + "mean_feedback_forward_cosine": 0.9997511640671761, + "mean_feedback_forward_relative_error": 0.021771032303091017, + "min_feedback_forward_cosine": 0.999232292175293 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.006626084240972996 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925308212637901, + "mean_feedback_forward_cosine": 0.9997511563762542, + "mean_feedback_forward_relative_error": 0.021770962003257968, + "min_feedback_forward_cosine": 0.999232292175293 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006670517491921783 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925303742289543, + "mean_feedback_forward_cosine": 0.999751175603559, + "mean_feedback_forward_relative_error": 0.021770884643398946, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0069802900040894745 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925300016999245, + "mean_feedback_forward_cosine": 0.9997511659899065, + "mean_feedback_forward_relative_error": 0.02177081188006747, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006564090129733085 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925294801592827, + "mean_feedback_forward_cosine": 0.9997511621444456, + "mean_feedback_forward_relative_error": 0.021770732687605966, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.007198657158911229 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925291821360588, + "mean_feedback_forward_cosine": 0.9997511775262894, + "mean_feedback_forward_relative_error": 0.021770659879210494, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006527402885854244 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0392528772354126, + "mean_feedback_forward_cosine": 0.9997511602217152, + "mean_feedback_forward_relative_error": 0.02177059237334517, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006383647970482707 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252836257219315, + "mean_feedback_forward_cosine": 0.999751171758098, + "mean_feedback_forward_relative_error": 0.021770518468392473, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0067850247360765935 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925279527902603, + "mean_feedback_forward_cosine": 0.9997511659899065, + "mean_feedback_forward_relative_error": 0.021770449295159307, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.005945792019739747 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252739399671555, + "mean_feedback_forward_cosine": 0.9997511813717503, + "mean_feedback_forward_relative_error": 0.02177037408335074, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.007371395756900311 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925269842147827, + "mean_feedback_forward_cosine": 0.9997511775262894, + "mean_feedback_forward_relative_error": 0.02177030530067221, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006610088579505682 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925265371799469, + "mean_feedback_forward_cosine": 0.9997511852172113, + "mean_feedback_forward_relative_error": 0.021770233543769966, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.006482609926313162 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925260901451111, + "mean_feedback_forward_cosine": 0.9997511794490199, + "mean_feedback_forward_relative_error": 0.021770161351249102, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006312023378238082 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252568036317825, + "mean_feedback_forward_cosine": 0.9997511775262894, + "mean_feedback_forward_relative_error": 0.021770090195200136, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006113427438139915 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925253078341484, + "mean_feedback_forward_cosine": 0.9997511948308637, + "mean_feedback_forward_relative_error": 0.021770016981228706, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006099065319225192 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925248980522156, + "mean_feedback_forward_cosine": 0.9997511640671761, + "mean_feedback_forward_relative_error": 0.021769947252206264, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006097415303662419 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925245255231857, + "mean_feedback_forward_cosine": 0.9997511602217152, + "mean_feedback_forward_relative_error": 0.021769877207735853, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.005813903130888939 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925241157412529, + "mean_feedback_forward_cosine": 0.999751175603559, + "mean_feedback_forward_relative_error": 0.021769803137548508, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006274959384575486 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252378046512604, + "mean_feedback_forward_cosine": 0.9997511659899065, + "mean_feedback_forward_relative_error": 0.021769732972907442, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.00568594042815268 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925233706831932, + "mean_feedback_forward_cosine": 0.9997511852172113, + "mean_feedback_forward_relative_error": 0.02176965893276276, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006060547957420349 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925229609012604, + "mean_feedback_forward_cosine": 0.9997511852172113, + "mean_feedback_forward_relative_error": 0.021769587866841786, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006110777246356011 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252255111932755, + "mean_feedback_forward_cosine": 0.9997511929081332, + "mean_feedback_forward_relative_error": 0.021769519489739212, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006131593030355871 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925221785902977, + "mean_feedback_forward_cosine": 0.9997511813717503, + "mean_feedback_forward_relative_error": 0.02176945082723133, + "min_feedback_forward_cosine": 0.9992325305938721 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0064618229238130155 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252180606126785, + "mean_feedback_forward_cosine": 0.9997511929081332, + "mean_feedback_forward_relative_error": 0.021769387962957545, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.005390891584306955 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0392521396279335, + "mean_feedback_forward_cosine": 0.9997511967535941, + "mean_feedback_forward_relative_error": 0.021769315169583404, + "min_feedback_forward_cosine": 0.9992323517799377 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006207142435461283 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925210237503052, + "mean_feedback_forward_cosine": 0.9997511909854028, + "mean_feedback_forward_relative_error": 0.021769242871913216, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005915933236479759 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039252061396837234, + "mean_feedback_forward_cosine": 0.9997511602217152, + "mean_feedback_forward_relative_error": 0.021769171655778925, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005669431590288878 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925202786922455, + "mean_feedback_forward_cosine": 0.9997511890626722, + "mean_feedback_forward_relative_error": 0.021769103158505693, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005420221343487501 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039251990616321564, + "mean_feedback_forward_cosine": 0.9997511813717503, + "mean_feedback_forward_relative_error": 0.02176903739511486, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.005751352581679821 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925195336341858, + "mean_feedback_forward_cosine": 0.9997511852172113, + "mean_feedback_forward_relative_error": 0.02176897052014547, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005610074019469321 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039251916110515594, + "mean_feedback_forward_cosine": 0.9997511890626722, + "mean_feedback_forward_relative_error": 0.021768901887680252, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.005644427023380995 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925187885761261, + "mean_feedback_forward_cosine": 0.999751204444516, + "mean_feedback_forward_relative_error": 0.021768835012710864, + "min_feedback_forward_cosine": 0.9992324113845825 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.005606720813140273 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03925183787941933, + "mean_feedback_forward_cosine": 0.9997511813717503, + "mean_feedback_forward_relative_error": 0.02176876387168323, + "min_feedback_forward_cosine": 0.9992324709892273 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0059318994602561 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039251793175935745, + "mean_feedback_forward_cosine": 0.9997512025217856, + "mean_feedback_forward_relative_error": 0.021768693061124895, + "min_feedback_forward_cosine": 0.9992325305938721 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0059052299538999795 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9209, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.4008916320800781, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2525847040, + "peak_memory_reserved_bytes": 3393191936, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.15913987159729, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10225.321983337402, + "train_wall_s": 10222.131490945816 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684202240000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2745656320000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json new file mode 100644 index 0000000..c25a9a7 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json @@ -0,0 +1,3390 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9996044099330902, + "feedback_forward_cosine": [ + 0.9998302459716797, + 0.9998152256011963, + 0.9996967315673828, + 0.9996758699417114, + 0.9997156262397766, + 0.9996615648269653, + 0.999754011631012, + 0.9996933341026306, + 0.9997844099998474, + 0.9996684789657593, + 0.9997125864028931, + 0.9997879266738892, + 0.9997588396072388, + 0.9997154474258423, + 0.9997693300247192, + 0.9997178316116333, + 0.9997802972793579, + 0.9997060298919678, + 0.9997598528862, + 0.9996607303619385, + 0.9997869729995728, + 0.9998668432235718, + 0.9998339414596558, + 0.9998186230659485, + 0.9998288154602051, + 0.9997850060462952, + 0.9998162984848022, + 0.9997274279594421, + 0.999793291091919, + 0.9995461106300354, + 0.9999839663505554 + ], + "feedback_forward_norm_ratio": [ + 0.9998724460601807, + 1.0005196332931519, + 0.9995152950286865, + 1.001641869544983, + 0.999545693397522, + 1.0011926889419556, + 0.9998229146003723, + 1.0016809701919556, + 0.9991549253463745, + 1.0004255771636963, + 0.9986014366149902, + 1.0009064674377441, + 0.999552845954895, + 1.0012096166610718, + 0.9988930225372314, + 1.0015215873718262, + 0.9993488788604736, + 1.001158356666565, + 0.9994612336158752, + 1.001421332359314, + 0.9993331432342529, + 1.000417709350586, + 0.9997079968452454, + 1.0008820295333862, + 0.9995848536491394, + 1.001044511795044, + 0.9995478987693787, + 1.001582384109497, + 0.9999879598617554, + 1.0011634826660156, + 1.0025697946548462 + ], + "feedback_forward_relative_error": [ + 0.018421653658151627, + 0.01923324540257454, + 0.024623440578579903, + 0.02553260512650013, + 0.0238448865711689, + 0.02605927363038063, + 0.022181058302521706, + 0.024846022948622704, + 0.020773712545633316, + 0.025759179145097733, + 0.024001268669962883, + 0.020627934485673904, + 0.02196275070309639, + 0.02390001341700554, + 0.021497350186109543, + 0.02382095344364643, + 0.02096632681787014, + 0.024285070598125458, + 0.02191520668566227, + 0.026104094460606575, + 0.020649723708629608, + 0.016331132501363754, + 0.01822110265493393, + 0.019073737785220146, + 0.018504101783037186, + 0.020774221047759056, + 0.019169632345438004, + 0.023417465388774872, + 0.020336169749498367, + 0.030168624594807625, + 0.006229218095541 + ], + "innovation_negative_gradient_cosine": [ + 0.9996152520179749, + 0.9995561242103577, + 0.9996229410171509, + 0.9995366930961609, + 0.9996248483657837, + 0.9995720386505127, + 0.9996442198753357, + 0.999592661857605, + 0.9996510744094849, + 0.9996282458305359, + 0.9996576905250549, + 0.9996254444122314, + 0.9996774792671204, + 0.9995726943016052, + 0.999693751335144, + 0.9995989799499512, + 0.9997098445892334, + 0.9996525645256042, + 0.9997299909591675, + 0.999668300151825, + 0.9997332692146301, + 0.9997650980949402, + 0.9997707009315491, + 0.9997556209564209, + 0.9997986555099487, + 0.9997751116752625, + 0.9998295307159424, + 0.9998553991317749, + 0.9998940229415894, + 0.9999567270278931, + 0.8186337351799011 + ], + "mean_feedback_forward_cosine": 0.9997565054124401, + "mean_feedback_forward_relative_error": 0.021717134742967544, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9996152520179749, + 0.9995561242103577, + 0.9996229410171509, + 0.9995366930961609, + 0.9996248483657837, + 0.9995720386505127, + 0.9996442198753357, + 0.999592661857605, + 0.9996510744094849, + 0.9996282458305359, + 0.9996576905250549, + 0.9996254444122314, + 0.9996774792671204, + 0.9995726943016052, + 0.999693751335144, + 0.9995989799499512, + 0.9997098445892334, + 0.9996525645256042, + 0.9997299909591675, + 0.999668300151825, + 0.9997332692146301, + 0.9997650980949402, + 0.9997707009315491, + 0.9997556209564209, + 0.9997986555099487, + 0.9997751116752625, + 0.9998295307159424, + 0.9998553991317749, + 0.9998940229415894, + 0.9999567270278931, + 0.8186337351799011 + ], + "teaching_negative_gradient_cosine": [ + 0.9996152520179749, + 0.9995561242103577, + 0.9996229410171509, + 0.9995366930961609, + 0.9996248483657837, + 0.9995720386505127, + 0.9996442198753357, + 0.999592661857605, + 0.9996510744094849, + 0.9996282458305359, + 0.9996576905250549, + 0.9996254444122314, + 0.9996774792671204, + 0.9995726943016052, + 0.999693751335144, + 0.9995989799499512, + 0.9997098445892334, + 0.9996525645256042, + 0.9997299909591675, + 0.999668300151825, + 0.9997332692146301, + 0.9997650980949402, + 0.9997707009315491, + 0.9997556209564209, + 0.9997986555099487, + 0.9997751116752625, + 0.9998295307159424, + 0.9998553991317749, + 0.9998940229415894, + 0.9999567270278931, + 0.8186337351799011 + ], + "wall_s": 0.11170530319213867 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3389378786087036, + "mean_feedback_forward_cosine": 0.23121365664466734, + "mean_feedback_forward_relative_error": 1.240019892492602, + "min_feedback_forward_cosine": 0.1128765344619751 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.9838821460723877 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2985464334487915, + "mean_feedback_forward_cosine": 0.32630072918630415, + "mean_feedback_forward_relative_error": 1.1610169679887834, + "min_feedback_forward_cosine": 0.16839854419231415 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.3787971196746827 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2493191957473755, + "mean_feedback_forward_cosine": 0.41258884005008206, + "mean_feedback_forward_relative_error": 1.0848919903078387, + "min_feedback_forward_cosine": 0.23476368188858032 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0900519537353515 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1967850923538208, + "mean_feedback_forward_cosine": 0.48254168033599854, + "mean_feedback_forward_relative_error": 1.0189406583386083, + "min_feedback_forward_cosine": 0.30199190974235535 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.906940527381897 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.160981297492981, + "mean_feedback_forward_cosine": 0.538169567623446, + "mean_feedback_forward_relative_error": 0.9634415920703642, + "min_feedback_forward_cosine": 0.3558718264102936 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.77196553440094 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1286283731460571, + "mean_feedback_forward_cosine": 0.5838505808384188, + "mean_feedback_forward_relative_error": 0.9149952463565334, + "min_feedback_forward_cosine": 0.39263737201690674 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6873360015869141 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0962344408035278, + "mean_feedback_forward_cosine": 0.6231732560742286, + "mean_feedback_forward_relative_error": 0.8710511913222652, + "min_feedback_forward_cosine": 0.4282245635986328 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6295898027706146 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0658029317855835, + "mean_feedback_forward_cosine": 0.658060630483012, + "mean_feedback_forward_relative_error": 0.8297388909324523, + "min_feedback_forward_cosine": 0.460544615983963 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5772971540260315 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0356392860412598, + "mean_feedback_forward_cosine": 0.6879042973441463, + "mean_feedback_forward_relative_error": 0.7924710483320297, + "min_feedback_forward_cosine": 0.4916229546070099 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.538321427745819 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0061553716659546, + "mean_feedback_forward_cosine": 0.7145854273150044, + "mean_feedback_forward_relative_error": 0.7575919608916005, + "min_feedback_forward_cosine": 0.5206334590911865 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5051441788768768 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.976460874080658, + "mean_feedback_forward_cosine": 0.738428444631638, + "mean_feedback_forward_relative_error": 0.7249820914960676, + "min_feedback_forward_cosine": 0.5487823486328125 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.48193352809906004 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9464977383613586, + "mean_feedback_forward_cosine": 0.7600205559884349, + "mean_feedback_forward_relative_error": 0.6940322133802599, + "min_feedback_forward_cosine": 0.5761566162109375 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4617367621898651 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9177645444869995, + "mean_feedback_forward_cosine": 0.779275276968556, + "mean_feedback_forward_relative_error": 0.6655332003870318, + "min_feedback_forward_cosine": 0.6014045476913452 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4340008737373352 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8895062208175659, + "mean_feedback_forward_cosine": 0.7972497728563124, + "mean_feedback_forward_relative_error": 0.6375727105525232, + "min_feedback_forward_cosine": 0.6256616115570068 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4249182174873352 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8613275289535522, + "mean_feedback_forward_cosine": 0.8134563046116983, + "mean_feedback_forward_relative_error": 0.6113178268555672, + "min_feedback_forward_cosine": 0.6489408016204834 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.40700346908569335 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8324474692344666, + "mean_feedback_forward_cosine": 0.8282331266710835, + "mean_feedback_forward_relative_error": 0.5862936689969032, + "min_feedback_forward_cosine": 0.6718162298202515 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.3925573239707947 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8057132363319397, + "mean_feedback_forward_cosine": 0.8414294950423702, + "mean_feedback_forward_relative_error": 0.5631513062023348, + "min_feedback_forward_cosine": 0.692278265953064 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.37800764578819274 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.780025064945221, + "mean_feedback_forward_cosine": 0.8532825554570844, + "mean_feedback_forward_relative_error": 0.5414041290360112, + "min_feedback_forward_cosine": 0.7115788459777832 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.35983910435676575 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7540488243103027, + "mean_feedback_forward_cosine": 0.8645286733104337, + "mean_feedback_forward_relative_error": 0.5199659630175559, + "min_feedback_forward_cosine": 0.7304697632789612 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3523705890464783 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7285679578781128, + "mean_feedback_forward_cosine": 0.8747497585511976, + "mean_feedback_forward_relative_error": 0.49976667088847004, + "min_feedback_forward_cosine": 0.7482064962387085 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3404514719581604 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7023472785949707, + "mean_feedback_forward_cosine": 0.8845681375072848, + "mean_feedback_forward_relative_error": 0.47956819447778887, + "min_feedback_forward_cosine": 0.7659227848052979 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3370584138774872 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6784579157829285, + "mean_feedback_forward_cosine": 0.8931465956472582, + "mean_feedback_forward_relative_error": 0.4611505793948327, + "min_feedback_forward_cosine": 0.7814069986343384 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32296759687423704 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6543031930923462, + "mean_feedback_forward_cosine": 0.9012511468702747, + "mean_feedback_forward_relative_error": 0.4430700550156255, + "min_feedback_forward_cosine": 0.7964164018630981 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.31611102496147153 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6313464641571045, + "mean_feedback_forward_cosine": 0.9085270089487876, + "mean_feedback_forward_relative_error": 0.4261899431866984, + "min_feedback_forward_cosine": 0.8101376295089722 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3068838739681244 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6094977855682373, + "mean_feedback_forward_cosine": 0.9153859999872023, + "mean_feedback_forward_relative_error": 0.4096834736485635, + "min_feedback_forward_cosine": 0.822838306427002 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3061916961765289 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5876675844192505, + "mean_feedback_forward_cosine": 0.9217286821334593, + "mean_feedback_forward_relative_error": 0.393863130961695, + "min_feedback_forward_cosine": 0.8350555300712585 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2910228369140625 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5656070113182068, + "mean_feedback_forward_cosine": 0.9276147388642834, + "mean_feedback_forward_relative_error": 0.37859375582587335, + "min_feedback_forward_cosine": 0.8469791412353516 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.28525240130901336 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5441515445709229, + "mean_feedback_forward_cosine": 0.9331032537644909, + "mean_feedback_forward_relative_error": 0.36385476204656786, + "min_feedback_forward_cosine": 0.8581732511520386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28182858699798585 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5238852500915527, + "mean_feedback_forward_cosine": 0.9381125915435052, + "mean_feedback_forward_relative_error": 0.34983458105594883, + "min_feedback_forward_cosine": 0.8683308362960815 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.27555020067214964 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5047932267189026, + "mean_feedback_forward_cosine": 0.9428141578551261, + "mean_feedback_forward_relative_error": 0.33617491443311015, + "min_feedback_forward_cosine": 0.8776401281356812 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2702294630479813 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48609381914138794, + "mean_feedback_forward_cosine": 0.9470732750431183, + "mean_feedback_forward_relative_error": 0.3233049591222117, + "min_feedback_forward_cosine": 0.8864260911941528 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26724182792663576 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46777960658073425, + "mean_feedback_forward_cosine": 0.9510572341180616, + "mean_feedback_forward_relative_error": 0.31077751204852133, + "min_feedback_forward_cosine": 0.8946413993835449 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2642540150499344 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4505240321159363, + "mean_feedback_forward_cosine": 0.954611559068003, + "mean_feedback_forward_relative_error": 0.29910426995446604, + "min_feedback_forward_cosine": 0.9021570086479187 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2528244213104248 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4330817759037018, + "mean_feedback_forward_cosine": 0.9580516219139099, + "mean_feedback_forward_relative_error": 0.28740497630450035, + "min_feedback_forward_cosine": 0.9094547033309937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2537894388771057 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4169538915157318, + "mean_feedback_forward_cosine": 0.9611637938407159, + "mean_feedback_forward_relative_error": 0.27640540441197736, + "min_feedback_forward_cosine": 0.9159486293792725 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2468545224380493 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4000706374645233, + "mean_feedback_forward_cosine": 0.9641662413074125, + "mean_feedback_forward_relative_error": 0.26542171835899353, + "min_feedback_forward_cosine": 0.9225339889526367 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24769609283447266 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3847455084323883, + "mean_feedback_forward_cosine": 0.9668317110307755, + "mean_feedback_forward_relative_error": 0.25530476531674784, + "min_feedback_forward_cosine": 0.9282823801040649 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24280201545238494 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36994194984436035, + "mean_feedback_forward_cosine": 0.9692680258904735, + "mean_feedback_forward_relative_error": 0.24568897389596508, + "min_feedback_forward_cosine": 0.9336087703704834 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.23785523980140685 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35572385787963867, + "mean_feedback_forward_cosine": 0.9715897787001825, + "mean_feedback_forward_relative_error": 0.2361768770121759, + "min_feedback_forward_cosine": 0.9385188817977905 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23314270420074462 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3419283628463745, + "mean_feedback_forward_cosine": 0.9737087891947839, + "mean_feedback_forward_relative_error": 0.22715740242312032, + "min_feedback_forward_cosine": 0.9431262016296387 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22954041654586793 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3289780020713806, + "mean_feedback_forward_cosine": 0.9756121846937365, + "mean_feedback_forward_relative_error": 0.21870481823721238, + "min_feedback_forward_cosine": 0.9472616910934448 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22527263911247253 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31575897336006165, + "mean_feedback_forward_cosine": 0.9774580540195588, + "mean_feedback_forward_relative_error": 0.21023723386949109, + "min_feedback_forward_cosine": 0.9514015316963196 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22636460013389587 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3028050661087036, + "mean_feedback_forward_cosine": 0.9791820549195812, + "mean_feedback_forward_relative_error": 0.201970340263459, + "min_feedback_forward_cosine": 0.9552354216575623 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2263233102607727 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2906534969806671, + "mean_feedback_forward_cosine": 0.980764873566166, + "mean_feedback_forward_relative_error": 0.1941150179313075, + "min_feedback_forward_cosine": 0.9587188959121704 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22183988755226136 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2791547179222107, + "mean_feedback_forward_cosine": 0.9821782688940724, + "mean_feedback_forward_relative_error": 0.1868205111353628, + "min_feedback_forward_cosine": 0.9618908166885376 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21705163217544557 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26807472109794617, + "mean_feedback_forward_cosine": 0.9835057720061271, + "mean_feedback_forward_relative_error": 0.17969327756474096, + "min_feedback_forward_cosine": 0.9648392200469971 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21400939811706543 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2570335268974304, + "mean_feedback_forward_cosine": 0.9847611227343159, + "mean_feedback_forward_relative_error": 0.17269634118964594, + "min_feedback_forward_cosine": 0.9676309823989868 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21496672592163085 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2468288391828537, + "mean_feedback_forward_cosine": 0.9858841088510328, + "mean_feedback_forward_relative_error": 0.1661904861609782, + "min_feedback_forward_cosine": 0.9701247811317444 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21069812502861024 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23741383850574493, + "mean_feedback_forward_cosine": 0.9869118948136607, + "mean_feedback_forward_relative_error": 0.15998160106039816, + "min_feedback_forward_cosine": 0.9723424911499023 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20564152919769288 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22796526551246643, + "mean_feedback_forward_cosine": 0.9878933891173332, + "mean_feedback_forward_relative_error": 0.15384173345181248, + "min_feedback_forward_cosine": 0.9744757413864136 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20788719606876374 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21927353739738464, + "mean_feedback_forward_cosine": 0.9887850899850169, + "mean_feedback_forward_relative_error": 0.14804859411331914, + "min_feedback_forward_cosine": 0.9763659238815308 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20169600876808166 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2106843739748001, + "mean_feedback_forward_cosine": 0.9896110103976342, + "mean_feedback_forward_relative_error": 0.14246293746175304, + "min_feedback_forward_cosine": 0.9781721234321594 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20488964691638947 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20236633718013763, + "mean_feedback_forward_cosine": 0.9903933617376512, + "mean_feedback_forward_relative_error": 0.13697626093222248, + "min_feedback_forward_cosine": 0.9798451066017151 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20032569314479828 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1942722648382187, + "mean_feedback_forward_cosine": 0.9911144433483001, + "mean_feedback_forward_relative_error": 0.13171462486347846, + "min_feedback_forward_cosine": 0.9814128875732422 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.1995404206466675 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18633601069450378, + "mean_feedback_forward_cosine": 0.9917949649595446, + "mean_feedback_forward_relative_error": 0.126550595967039, + "min_feedback_forward_cosine": 0.9828876256942749 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.20286603238105774 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17927993834018707, + "mean_feedback_forward_cosine": 0.9923889463947665, + "mean_feedback_forward_relative_error": 0.12187032990397946, + "min_feedback_forward_cosine": 0.9841449856758118 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19383209038257598 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17214931547641754, + "mean_feedback_forward_cosine": 0.9929645119174835, + "mean_feedback_forward_relative_error": 0.11715754241712632, + "min_feedback_forward_cosine": 0.9853746891021729 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1954141134405136 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1658857762813568, + "mean_feedback_forward_cosine": 0.9934704149923017, + "mean_feedback_forward_relative_error": 0.11285452989320602, + "min_feedback_forward_cosine": 0.9864099621772766 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19057123153686523 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15934672951698303, + "mean_feedback_forward_cosine": 0.9939606555046574, + "mean_feedback_forward_relative_error": 0.10851789398058768, + "min_feedback_forward_cosine": 0.9874546527862549 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.1898304710149765 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15314742922782898, + "mean_feedback_forward_cosine": 0.9944075519038785, + "mean_feedback_forward_relative_error": 0.10440802309782274, + "min_feedback_forward_cosine": 0.9884060621261597 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19179069291591644 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14699344336986542, + "mean_feedback_forward_cosine": 0.9948373225427443, + "mean_feedback_forward_relative_error": 0.1003009861996097, + "min_feedback_forward_cosine": 0.9893105030059814 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1871666722202301 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1410326063632965, + "mean_feedback_forward_cosine": 0.9952297114556835, + "mean_feedback_forward_relative_error": 0.09641332095188479, + "min_feedback_forward_cosine": 0.9901542067527771 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18810378356218338 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13551799952983856, + "mean_feedback_forward_cosine": 0.9955768777478126, + "mean_feedback_forward_relative_error": 0.09282779801757104, + "min_feedback_forward_cosine": 0.9909045696258545 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1841782896709442 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12996059656143188, + "mean_feedback_forward_cosine": 0.9959184700442899, + "mean_feedback_forward_relative_error": 0.08917618771233866, + "min_feedback_forward_cosine": 0.991633415222168 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1905988346195221 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12496453523635864, + "mean_feedback_forward_cosine": 0.9962086273777869, + "mean_feedback_forward_relative_error": 0.0859366656551438, + "min_feedback_forward_cosine": 0.9922622442245483 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18150609948158264 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1199355497956276, + "mean_feedback_forward_cosine": 0.9964944797177469, + "mean_feedback_forward_relative_error": 0.08262782597974423, + "min_feedback_forward_cosine": 0.9928679466247559 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18521738644123079 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11520607024431229, + "mean_feedback_forward_cosine": 0.9967465746787286, + "mean_feedback_forward_relative_error": 0.07959193477947865, + "min_feedback_forward_cosine": 0.9934182167053223 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1773731155395508 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11053729802370071, + "mean_feedback_forward_cosine": 0.9969997386778554, + "mean_feedback_forward_relative_error": 0.07642521360708822, + "min_feedback_forward_cosine": 0.9939394593238831 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18311363577842713 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10619539022445679, + "mean_feedback_forward_cosine": 0.9972271823113964, + "mean_feedback_forward_relative_error": 0.0734694192726766, + "min_feedback_forward_cosine": 0.9944032430648804 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18016735112667084 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10208764672279358, + "mean_feedback_forward_cosine": 0.9974322626667638, + "mean_feedback_forward_relative_error": 0.07069606442124612, + "min_feedback_forward_cosine": 0.9948270916938782 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1764145338010788 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09786710888147354, + "mean_feedback_forward_cosine": 0.9976302096920628, + "mean_feedback_forward_relative_error": 0.06791745264443659, + "min_feedback_forward_cosine": 0.9952437877655029 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18184925278186798 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09394015371799469, + "mean_feedback_forward_cosine": 0.9978082237705108, + "mean_feedback_forward_relative_error": 0.06531861268224255, + "min_feedback_forward_cosine": 0.9956168532371521 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.17577416330337525 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09005498141050339, + "mean_feedback_forward_cosine": 0.9979749706483656, + "mean_feedback_forward_relative_error": 0.06278794499174241, + "min_feedback_forward_cosine": 0.9959706664085388 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1818134422826767 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08653854578733444, + "mean_feedback_forward_cosine": 0.998121888406815, + "mean_feedback_forward_relative_error": 0.06045928212904161, + "min_feedback_forward_cosine": 0.9962779879570007 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17373866063117982 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0829930230975151, + "mean_feedback_forward_cosine": 0.9982667949891859, + "mean_feedback_forward_relative_error": 0.058078636325174765, + "min_feedback_forward_cosine": 0.9965764880180359 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17920592172145844 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07985182106494904, + "mean_feedback_forward_cosine": 0.9983929434130269, + "mean_feedback_forward_relative_error": 0.055919353279375264, + "min_feedback_forward_cosine": 0.9968297481536865 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17177996276378632 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07659643143415451, + "mean_feedback_forward_cosine": 0.9985156078492442, + "mean_feedback_forward_relative_error": 0.05374190770089626, + "min_feedback_forward_cosine": 0.9970822334289551 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17647761666297912 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07359126955270767, + "mean_feedback_forward_cosine": 0.9986273915536942, + "mean_feedback_forward_relative_error": 0.05167639658095375, + "min_feedback_forward_cosine": 0.997306227684021 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17503470136404037 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07059352844953537, + "mean_feedback_forward_cosine": 0.9987327283428561, + "mean_feedback_forward_relative_error": 0.04965663653227591, + "min_feedback_forward_cosine": 0.9975208044052124 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1740077546787262 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06786178797483444, + "mean_feedback_forward_cosine": 0.9988272247775909, + "mean_feedback_forward_relative_error": 0.04776773012934193, + "min_feedback_forward_cosine": 0.9977082014083862 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.16993277304172516 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06507941335439682, + "mean_feedback_forward_cosine": 0.9989168317087235, + "mean_feedback_forward_relative_error": 0.04590913677407849, + "min_feedback_forward_cosine": 0.9978917837142944 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17373484318494797 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0625593364238739, + "mean_feedback_forward_cosine": 0.9989959405314538, + "mean_feedback_forward_relative_error": 0.044198656394597025, + "min_feedback_forward_cosine": 0.9980515837669373 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16842556659698488 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.060116615146398544, + "mean_feedback_forward_cosine": 0.9990705097875288, + "mean_feedback_forward_relative_error": 0.042526453883657535, + "min_feedback_forward_cosine": 0.9982008934020996 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1669404342365265 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.057767800986766815, + "mean_feedback_forward_cosine": 0.9991391704928491, + "mean_feedback_forward_relative_error": 0.04092263841941472, + "min_feedback_forward_cosine": 0.9983383417129517 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16348597554206848 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055436767637729645, + "mean_feedback_forward_cosine": 0.9992048394295477, + "mean_feedback_forward_relative_error": 0.0393322111377793, + "min_feedback_forward_cosine": 0.9984695911407471 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.17286750542640686 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05322682857513428, + "mean_feedback_forward_cosine": 0.9992634538681276, + "mean_feedback_forward_relative_error": 0.037855942883799155, + "min_feedback_forward_cosine": 0.9985889196395874 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16786843843221663 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05116841569542885, + "mean_feedback_forward_cosine": 0.9993203743811576, + "mean_feedback_forward_relative_error": 0.03636146069414193, + "min_feedback_forward_cosine": 0.998695969581604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16632413774490357 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.049093734472990036, + "mean_feedback_forward_cosine": 0.9993729649051544, + "mean_feedback_forward_relative_error": 0.03492417012251193, + "min_feedback_forward_cosine": 0.9987992644309998 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16653296973705292 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.047124844044446945, + "mean_feedback_forward_cosine": 0.9994193950007039, + "mean_feedback_forward_relative_error": 0.0336092067581992, + "min_feedback_forward_cosine": 0.9988933801651001 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16568008406162263 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04536622390151024, + "mean_feedback_forward_cosine": 0.99946129322052, + "mean_feedback_forward_relative_error": 0.03237246349453926, + "min_feedback_forward_cosine": 0.9989744424819946 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16156356994628907 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04356296733021736, + "mean_feedback_forward_cosine": 0.9995024434981807, + "mean_feedback_forward_relative_error": 0.03111092064289316, + "min_feedback_forward_cosine": 0.9990541338920593 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16308865074634551 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0419180728495121, + "mean_feedback_forward_cosine": 0.9995387408041185, + "mean_feedback_forward_relative_error": 0.029956305850177042, + "min_feedback_forward_cosine": 0.9991241693496704 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15842433814048767 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04021833837032318, + "mean_feedback_forward_cosine": 0.9995746266457343, + "mean_feedback_forward_relative_error": 0.028767979643758265, + "min_feedback_forward_cosine": 0.9991938471794128 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16590015714645387 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03862576559185982, + "mean_feedback_forward_cosine": 0.9996064747533491, + "mean_feedback_forward_relative_error": 0.02766890376205406, + "min_feedback_forward_cosine": 0.999256432056427 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16425485748291016 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03712113946676254, + "mean_feedback_forward_cosine": 0.999635394542448, + "mean_feedback_forward_relative_error": 0.026631538575935747, + "min_feedback_forward_cosine": 0.9993130564689636 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16208505117416383 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03564485162496567, + "mean_feedback_forward_cosine": 0.9996625915650399, + "mean_feedback_forward_relative_error": 0.02561724678643288, + "min_feedback_forward_cosine": 0.9993666410446167 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16008623097419739 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03427170217037201, + "mean_feedback_forward_cosine": 0.999687377483614, + "mean_feedback_forward_relative_error": 0.024657781386087017, + "min_feedback_forward_cosine": 0.9994142651557922 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1584984017133713 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03291971608996391, + "mean_feedback_forward_cosine": 0.9997108097999327, + "mean_feedback_forward_relative_error": 0.0237135432900921, + "min_feedback_forward_cosine": 0.9994597434997559 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15919874101161957 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03161061182618141, + "mean_feedback_forward_cosine": 0.9997326616317995, + "mean_feedback_forward_relative_error": 0.02280093639368011, + "min_feedback_forward_cosine": 0.9995017647743225 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16238198031425477 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030365271493792534, + "mean_feedback_forward_cosine": 0.999752281173583, + "mean_feedback_forward_relative_error": 0.021947178478923537, + "min_feedback_forward_cosine": 0.9995402097702026 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1577615320110321 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03035474568605423, + "mean_feedback_forward_cosine": 0.9997524522965954, + "mean_feedback_forward_relative_error": 0.02193744862151723, + "min_feedback_forward_cosine": 0.9995405673980713 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09112625179052353 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03034961223602295, + "mean_feedback_forward_cosine": 0.9997525426649279, + "mean_feedback_forward_relative_error": 0.02193117661461715, + "min_feedback_forward_cosine": 0.9995406270027161 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.0651662121462822 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030345071107149124, + "mean_feedback_forward_cosine": 0.9997526368787212, + "mean_feedback_forward_relative_error": 0.021925508570406708, + "min_feedback_forward_cosine": 0.9995408654212952 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05518694655299187 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030340731143951416, + "mean_feedback_forward_cosine": 0.9997527310925145, + "mean_feedback_forward_relative_error": 0.02192018811981524, + "min_feedback_forward_cosine": 0.9995409250259399 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.05016247516512871 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03033665381371975, + "mean_feedback_forward_cosine": 0.999752809924464, + "mean_feedback_forward_relative_error": 0.02191504654324343, + "min_feedback_forward_cosine": 0.9995411038398743 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.045509788563251496 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0303326603025198, + "mean_feedback_forward_cosine": 0.9997529002927965, + "mean_feedback_forward_relative_error": 0.021910079619697025, + "min_feedback_forward_cosine": 0.999541163444519 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04068453124523163 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030328577384352684, + "mean_feedback_forward_cosine": 0.9997529579747108, + "mean_feedback_forward_relative_error": 0.02190520210311778, + "min_feedback_forward_cosine": 0.9995412826538086 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03963619037270546 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030324583873152733, + "mean_feedback_forward_cosine": 0.9997530579566956, + "mean_feedback_forward_relative_error": 0.02190041650206812, + "min_feedback_forward_cosine": 0.9995415210723877 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03669267584860325 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030320720747113228, + "mean_feedback_forward_cosine": 0.9997531310204537, + "mean_feedback_forward_relative_error": 0.02189569534253209, + "min_feedback_forward_cosine": 0.9995415210723877 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03461138196527958 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030317235738039017, + "mean_feedback_forward_cosine": 0.9997532002387508, + "mean_feedback_forward_relative_error": 0.021891336106965618, + "min_feedback_forward_cosine": 0.9995415806770325 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03089582992076874 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03031330741941929, + "mean_feedback_forward_cosine": 0.9997532829161613, + "mean_feedback_forward_relative_error": 0.021886823428494316, + "min_feedback_forward_cosine": 0.9995417594909668 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02949810125529766 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030309543013572693, + "mean_feedback_forward_cosine": 0.9997533540571889, + "mean_feedback_forward_relative_error": 0.021882356520021153, + "min_feedback_forward_cosine": 0.9995419979095459 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.02843426930308342 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030305659398436546, + "mean_feedback_forward_cosine": 0.9997534617300956, + "mean_feedback_forward_relative_error": 0.021877802427737944, + "min_feedback_forward_cosine": 0.9995419979095459 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.027714148516058922 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030302129685878754, + "mean_feedback_forward_cosine": 0.9997535290256623, + "mean_feedback_forward_relative_error": 0.021873503232434872, + "min_feedback_forward_cosine": 0.9995420575141907 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02623457520365715 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030298389494419098, + "mean_feedback_forward_cosine": 0.9997536097803423, + "mean_feedback_forward_relative_error": 0.021869151507534327, + "min_feedback_forward_cosine": 0.9995421767234802 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.024785830131173133 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03029458038508892, + "mean_feedback_forward_cosine": 0.9997536751531786, + "mean_feedback_forward_relative_error": 0.021864845177098628, + "min_feedback_forward_cosine": 0.9995423555374146 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024115232612490652 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03029109351336956, + "mean_feedback_forward_cosine": 0.9997537635987804, + "mean_feedback_forward_relative_error": 0.021860653369296944, + "min_feedback_forward_cosine": 0.9995425939559937 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.022689599386453627 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0302873644977808, + "mean_feedback_forward_cosine": 0.9997538347398082, + "mean_feedback_forward_relative_error": 0.021856272938631235, + "min_feedback_forward_cosine": 0.9995425343513489 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02323253556996584 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03028363548219204, + "mean_feedback_forward_cosine": 0.9997539212626796, + "mean_feedback_forward_relative_error": 0.02185203465482881, + "min_feedback_forward_cosine": 0.9995427131652832 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.02146149538576603 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030279796570539474, + "mean_feedback_forward_cosine": 0.9997540058628205, + "mean_feedback_forward_relative_error": 0.021847640765049765, + "min_feedback_forward_cosine": 0.9995428323745728 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.021298135353326798 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0302763469517231, + "mean_feedback_forward_cosine": 0.9997540866175005, + "mean_feedback_forward_relative_error": 0.021843459547287034, + "min_feedback_forward_cosine": 0.9995429515838623 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.019140538485646248 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030272936448454857, + "mean_feedback_forward_cosine": 0.9997541308403015, + "mean_feedback_forward_relative_error": 0.021839403006578644, + "min_feedback_forward_cosine": 0.9995430111885071 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01843281641423702 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030269691720604897, + "mean_feedback_forward_cosine": 0.999754253895052, + "mean_feedback_forward_relative_error": 0.021835425523139777, + "min_feedback_forward_cosine": 0.9995430707931519 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.017320715464651585 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030265865847468376, + "mean_feedback_forward_cosine": 0.9997542846587396, + "mean_feedback_forward_relative_error": 0.02183128496812236, + "min_feedback_forward_cosine": 0.999543309211731 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.018616918771862983 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03026205487549305, + "mean_feedback_forward_cosine": 0.9997543692588806, + "mean_feedback_forward_relative_error": 0.021826918447209943, + "min_feedback_forward_cosine": 0.999543309211731 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01787962683171034 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03025832027196884, + "mean_feedback_forward_cosine": 0.9997544730863264, + "mean_feedback_forward_relative_error": 0.021822680538940813, + "min_feedback_forward_cosine": 0.9995434880256653 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.016799339402318002 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030255036428570747, + "mean_feedback_forward_cosine": 0.9997545269227797, + "mean_feedback_forward_relative_error": 0.021818770065663323, + "min_feedback_forward_cosine": 0.9995435476303101 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.015555499247312546 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030251365154981613, + "mean_feedback_forward_cosine": 0.9997546096001902, + "mean_feedback_forward_relative_error": 0.02181462037767614, + "min_feedback_forward_cosine": 0.9995437264442444 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01577266572535038 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030247338116168976, + "mean_feedback_forward_cosine": 0.9997547076594445, + "mean_feedback_forward_relative_error": 0.021810064212449135, + "min_feedback_forward_cosine": 0.9995439052581787 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.016338454892113805 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030243637040257454, + "mean_feedback_forward_cosine": 0.9997547807232026, + "mean_feedback_forward_relative_error": 0.021805861138648564, + "min_feedback_forward_cosine": 0.9995437860488892 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.015070238414108754 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030239706858992577, + "mean_feedback_forward_cosine": 0.999754878782457, + "mean_feedback_forward_relative_error": 0.02180133913193018, + "min_feedback_forward_cosine": 0.9995439052581787 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.015652655007243155 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030236007645726204, + "mean_feedback_forward_cosine": 0.9997549729962503, + "mean_feedback_forward_relative_error": 0.021797153978578506, + "min_feedback_forward_cosine": 0.9995441436767578 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.014617297267764806 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030232984572649002, + "mean_feedback_forward_cosine": 0.9997550345236256, + "mean_feedback_forward_relative_error": 0.021793444460678484, + "min_feedback_forward_cosine": 0.9995443224906921 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013100630081892014 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030228734016418457, + "mean_feedback_forward_cosine": 0.9997551095101141, + "mean_feedback_forward_relative_error": 0.02178893143671655, + "min_feedback_forward_cosine": 0.9995443224906921 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.014554911104142665 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030225105583667755, + "mean_feedback_forward_cosine": 0.9997551729602199, + "mean_feedback_forward_relative_error": 0.021784932487794468, + "min_feedback_forward_cosine": 0.9995445013046265 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.012763612243533135 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030221687629818916, + "mean_feedback_forward_cosine": 0.9997552633285522, + "mean_feedback_forward_relative_error": 0.021780829591255997, + "min_feedback_forward_cosine": 0.9995444416999817 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.012405773229599 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030217928811907768, + "mean_feedback_forward_cosine": 0.999755367155998, + "mean_feedback_forward_relative_error": 0.021776424675819375, + "min_feedback_forward_cosine": 0.9995446801185608 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.013631196454167367 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030214227735996246, + "mean_feedback_forward_cosine": 0.9997554498334085, + "mean_feedback_forward_relative_error": 0.021772182321236018, + "min_feedback_forward_cosine": 0.9995447397232056 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012717415338754654 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030210770666599274, + "mean_feedback_forward_cosine": 0.999755528665358, + "mean_feedback_forward_relative_error": 0.02176817787450648, + "min_feedback_forward_cosine": 0.9995449185371399 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.011400041757076979 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030206751078367233, + "mean_feedback_forward_cosine": 0.99975561711096, + "mean_feedback_forward_relative_error": 0.021763561578888084, + "min_feedback_forward_cosine": 0.9995450377464294 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012817322442829609 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03020297735929489, + "mean_feedback_forward_cosine": 0.9997557074792923, + "mean_feedback_forward_relative_error": 0.02175934123055589, + "min_feedback_forward_cosine": 0.999545156955719 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.01158828136779368 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030199110507965088, + "mean_feedback_forward_cosine": 0.999755774774859, + "mean_feedback_forward_relative_error": 0.021755064341930613, + "min_feedback_forward_cosine": 0.9995452761650085 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011865053795874119 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030195636674761772, + "mean_feedback_forward_cosine": 0.9997558612977305, + "mean_feedback_forward_relative_error": 0.021750994792748845, + "min_feedback_forward_cosine": 0.9995453357696533 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.010737874366939067 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030192187055945396, + "mean_feedback_forward_cosine": 0.9997559285932972, + "mean_feedback_forward_relative_error": 0.021747009242854772, + "min_feedback_forward_cosine": 0.9995453357696533 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.010735899466872215 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030188128352165222, + "mean_feedback_forward_cosine": 0.9997560497253172, + "mean_feedback_forward_relative_error": 0.02174242093197761, + "min_feedback_forward_cosine": 0.9995455741882324 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.01134729971215129 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03018438071012497, + "mean_feedback_forward_cosine": 0.9997561420163801, + "mean_feedback_forward_relative_error": 0.021738111041486263, + "min_feedback_forward_cosine": 0.9995457530021667 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010523225945830345 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030180975794792175, + "mean_feedback_forward_cosine": 0.9997562112346772, + "mean_feedback_forward_relative_error": 0.02173394923128428, + "min_feedback_forward_cosine": 0.999545693397522 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.009820032843351364 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03017698973417282, + "mean_feedback_forward_cosine": 0.999756278530244, + "mean_feedback_forward_relative_error": 0.021729366147830602, + "min_feedback_forward_cosine": 0.9995459318161011 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.010756156692318618 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030173545703291893, + "mean_feedback_forward_cosine": 0.9997563862031505, + "mean_feedback_forward_relative_error": 0.02172539821795879, + "min_feedback_forward_cosine": 0.9995459914207458 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.009835685213804245 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169757083058357, + "mean_feedback_forward_cosine": 0.9997564708032916, + "mean_feedback_forward_relative_error": 0.021721043413685213, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009867143030613661 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169716104865074, + "mean_feedback_forward_cosine": 0.9997564631123697, + "mean_feedback_forward_relative_error": 0.021720950416619738, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008295226010084152 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016967698931694, + "mean_feedback_forward_cosine": 0.9997564746487525, + "mean_feedback_forward_relative_error": 0.021720858801516793, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.00862344180509448 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016965463757515, + "mean_feedback_forward_cosine": 0.9997564631123697, + "mean_feedback_forward_relative_error": 0.021720769514720285, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008312926781252027 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016963228583336, + "mean_feedback_forward_cosine": 0.9997564708032916, + "mean_feedback_forward_relative_error": 0.021720683127040823, + "min_feedback_forward_cosine": 0.9995459914207458 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.007363148895800114 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016960807144642, + "mean_feedback_forward_cosine": 0.9997564669578306, + "mean_feedback_forward_relative_error": 0.021720602297254147, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.00785008468285203 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169574543833733, + "mean_feedback_forward_cosine": 0.9997564823396744, + "mean_feedback_forward_relative_error": 0.0217205208215502, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007802247190997005 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169550329446793, + "mean_feedback_forward_cosine": 0.9997564727260221, + "mean_feedback_forward_relative_error": 0.02172043769349975, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.0067719343725591895 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169522389769554, + "mean_feedback_forward_cosine": 0.9997564746487525, + "mean_feedback_forward_relative_error": 0.02172035832078226, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.006957013593614101 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169496312737465, + "mean_feedback_forward_cosine": 0.999756476571483, + "mean_feedback_forward_relative_error": 0.021720276709886327, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.006896061420664191 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169475823640823, + "mean_feedback_forward_cosine": 0.9997564746487525, + "mean_feedback_forward_relative_error": 0.02172020019122189, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007256742753982544 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016945719718933, + "mean_feedback_forward_cosine": 0.9997564881078659, + "mean_feedback_forward_relative_error": 0.02172011966186185, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007564757667630911 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169425532221794, + "mean_feedback_forward_cosine": 0.999756480416944, + "mean_feedback_forward_relative_error": 0.02172003458103826, + "min_feedback_forward_cosine": 0.999546229839325 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007529743331968784 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169401317834854, + "mean_feedback_forward_cosine": 0.9997564554214478, + "mean_feedback_forward_relative_error": 0.021719952128947742, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007267204804420471 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169377103447914, + "mean_feedback_forward_cosine": 0.999756480416944, + "mean_feedback_forward_relative_error": 0.02171987280129425, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006840419639255852 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169356614351273, + "mean_feedback_forward_cosine": 0.9997564784942135, + "mean_feedback_forward_relative_error": 0.021719793128150124, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0072005226527899506 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169332399964333, + "mean_feedback_forward_cosine": 0.999756476571483, + "mean_feedback_forward_relative_error": 0.02171971240351277, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.006555082276463509 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016931563615799, + "mean_feedback_forward_cosine": 0.9997564900305963, + "mean_feedback_forward_relative_error": 0.021719636125189644, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0067100419231597335 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169295147061348, + "mean_feedback_forward_cosine": 0.9997564727260221, + "mean_feedback_forward_relative_error": 0.021719558284648003, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006859775168448687 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169276520609856, + "mean_feedback_forward_cosine": 0.9997564727260221, + "mean_feedback_forward_relative_error": 0.02171948000848774, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.007023929670471698 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169248580932617, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.021719402137903437, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.00637954917833209 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169226229190826, + "mean_feedback_forward_cosine": 0.9997564881078659, + "mean_feedback_forward_relative_error": 0.021719321548458067, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006854683260917664 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169205740094185, + "mean_feedback_forward_cosine": 0.9997564842624049, + "mean_feedback_forward_relative_error": 0.021719245540518916, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0066349893228709695 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169187113642693, + "mean_feedback_forward_cosine": 0.9997565015669791, + "mean_feedback_forward_relative_error": 0.021719168976790482, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006290283277481794 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169174075126648, + "mean_feedback_forward_cosine": 0.9997564957987878, + "mean_feedback_forward_relative_error": 0.021719093764981916, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006629161743670702 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169157311320305, + "mean_feedback_forward_cosine": 0.9997564881078659, + "mean_feedback_forward_relative_error": 0.021719019078919963, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006182823036164045 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169138684868813, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.021718938955135884, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006323601663410664 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169112607836723, + "mean_feedback_forward_cosine": 0.9997565054124401, + "mean_feedback_forward_relative_error": 0.021718856503045367, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006986099981889129 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169080942869186, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.0217187755831307, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.006972612196803093 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169058591127396, + "mean_feedback_forward_cosine": 0.999756476571483, + "mean_feedback_forward_relative_error": 0.021718700311236804, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0067703842197358605 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169036239385605, + "mean_feedback_forward_cosine": 0.999756513103362, + "mean_feedback_forward_relative_error": 0.021718626541476094, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0063818694199621676 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030169015750288963, + "mean_feedback_forward_cosine": 0.9997564784942135, + "mean_feedback_forward_relative_error": 0.021718546733139985, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006102044712156057 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016900084912777, + "mean_feedback_forward_cosine": 0.9997564938760573, + "mean_feedback_forward_relative_error": 0.021718474240192482, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.005858240797370672 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016897849738598, + "mean_feedback_forward_cosine": 0.9997565034897097, + "mean_feedback_forward_relative_error": 0.02171839940391721, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.005866434761993587 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168967321515083, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.021718321923887538, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.005773927080184221 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168946832418442, + "mean_feedback_forward_cosine": 0.9997564996442487, + "mean_feedback_forward_relative_error": 0.021718249115492066, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0064538452906906605 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168922618031502, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.021718173317851558, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006007886548340321 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168896540999413, + "mean_feedback_forward_cosine": 0.9997564977215182, + "mean_feedback_forward_relative_error": 0.021718097685445702, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.005968254810981452 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016887977719307, + "mean_feedback_forward_cosine": 0.999756513103362, + "mean_feedback_forward_relative_error": 0.02171802355517303, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.00620140129506588 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016885556280613, + "mean_feedback_forward_cosine": 0.9997564919533268, + "mean_feedback_forward_relative_error": 0.021717947877703175, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.005907446707040072 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016883134841919, + "mean_feedback_forward_cosine": 0.999756513103362, + "mean_feedback_forward_relative_error": 0.02171787343198253, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.005722253931760788 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168810859322548, + "mean_feedback_forward_cosine": 0.999756513103362, + "mean_feedback_forward_relative_error": 0.02171779963217916, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.005939775842446834 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168788507580757, + "mean_feedback_forward_cosine": 0.9997565034897097, + "mean_feedback_forward_relative_error": 0.021717723173600052, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006015781491547823 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016876056790352, + "mean_feedback_forward_cosine": 0.9997564977215182, + "mean_feedback_forward_relative_error": 0.02171765173214578, + "min_feedback_forward_cosine": 0.9995459914207458 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005618943987675011 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016873449087143, + "mean_feedback_forward_cosine": 0.9997564957987878, + "mean_feedback_forward_relative_error": 0.021717575558971976, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.006377606983333826 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168725177645683, + "mean_feedback_forward_cosine": 0.9997565073351706, + "mean_feedback_forward_relative_error": 0.02171750329134445, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.005670744403153658 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03016870841383934, + "mean_feedback_forward_cosine": 0.9997564861851354, + "mean_feedback_forward_relative_error": 0.021717430452906316, + "min_feedback_forward_cosine": 0.9995461702346802 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005683699643909931 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168689787387848, + "mean_feedback_forward_cosine": 0.9997565054124401, + "mean_feedback_forward_relative_error": 0.02171735931187868, + "min_feedback_forward_cosine": 0.9995459914207458 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0059974849678576 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168669298291206, + "mean_feedback_forward_cosine": 0.9997564957987878, + "mean_feedback_forward_relative_error": 0.021717282222403635, + "min_feedback_forward_cosine": 0.9995460510253906 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.006048647853694856 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168643221259117, + "mean_feedback_forward_cosine": 0.9997565188715535, + "mean_feedback_forward_relative_error": 0.021717206334635135, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005997108866162598 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.030168624594807625, + "mean_feedback_forward_cosine": 0.9997565054124401, + "mean_feedback_forward_relative_error": 0.021717134742967544, + "min_feedback_forward_cosine": 0.9995461106300354 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005753214195370674 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9234, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3684505432128906, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2525847040, + "peak_memory_reserved_bytes": 3393191936, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1665029525756836, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10229.90122294426, + "train_wall_s": 10226.761625766754 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684202240000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2745656320000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json new file mode 100644 index 0000000..c2f5072 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json @@ -0,0 +1,3390 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9996268689632416, + "feedback_forward_cosine": [ + 0.9997919797897339, + 0.9997849464416504, + 0.9997409582138062, + 0.9997344613075256, + 0.999788224697113, + 0.9997414350509644, + 0.9998248815536499, + 0.9997788667678833, + 0.9997860789299011, + 0.9996800422668457, + 0.9997262954711914, + 0.9998047351837158, + 0.9997510313987732, + 0.9997111558914185, + 0.9997680187225342, + 0.999701976776123, + 0.9997506737709045, + 0.9996946454048157, + 0.999782919883728, + 0.9996932744979858, + 0.9997825622558594, + 0.9998618364334106, + 0.99983149766922, + 0.9998146295547485, + 0.9998298287391663, + 0.9997783899307251, + 0.9998220205307007, + 0.9996968507766724, + 0.9997162222862244, + 0.9994392395019531, + 0.9999843835830688 + ], + "feedback_forward_norm_ratio": [ + 0.9989004731178284, + 1.0007294416427612, + 1.0002119541168213, + 1.001495122909546, + 1.0001250505447388, + 1.0014092922210693, + 0.9999539256095886, + 1.0007349252700806, + 0.9999049305915833, + 1.0011711120605469, + 0.9987809658050537, + 1.001015543937683, + 0.9990402460098267, + 1.0010043382644653, + 0.999497652053833, + 1.0015132427215576, + 0.9995678663253784, + 1.0010426044464111, + 0.9991808533668518, + 1.0010385513305664, + 0.999531626701355, + 1.0005073547363281, + 0.9997082352638245, + 1.000762939453125, + 0.9995765089988708, + 1.000929594039917, + 0.9997575879096985, + 1.0009090900421143, + 0.9997023940086365, + 1.0010747909545898, + 1.0025800466537476 + ], + "feedback_forward_relative_error": [ + 0.020418580621480942, + 0.020756781101226807, + 0.022764239460229874, + 0.023108825087547302, + 0.020585954189300537, + 0.022801287472248077, + 0.018710697069764137, + 0.021051138639450073, + 0.020687395706772804, + 0.025339433923363686, + 0.02341407909989357, + 0.019799573346972466, + 0.022325001657009125, + 0.02407108061015606, + 0.02154187485575676, + 0.024480070918798447, + 0.02232399955391884, + 0.02474418468773365, + 0.02083893120288849, + 0.024802938103675842, + 0.020856020972132683, + 0.016635606065392494, + 0.018359307199716568, + 0.019281823188066483, + 0.018446631729602814, + 0.021085331216454506, + 0.01886802539229393, + 0.024646619334816933, + 0.023822002112865448, + 0.03352376073598862, + 0.006160619668662548 + ], + "innovation_negative_gradient_cosine": [ + 0.999638557434082, + 0.9996102452278137, + 0.9996362924575806, + 0.9995453357696533, + 0.9996436834335327, + 0.9995719194412231, + 0.9996561408042908, + 0.999657154083252, + 0.999669075012207, + 0.9996402859687805, + 0.9996742010116577, + 0.9996456503868103, + 0.999691903591156, + 0.9995970129966736, + 0.9997073411941528, + 0.9996359348297119, + 0.9997234344482422, + 0.999672532081604, + 0.9997422695159912, + 0.9996891617774963, + 0.9997507333755493, + 0.9997844696044922, + 0.9997909069061279, + 0.9997697472572327, + 0.9998173713684082, + 0.9998141527175903, + 0.9998586177825928, + 0.9998947381973267, + 0.9999035596847534, + 0.9999581575393677, + 0.7821329832077026 + ], + "mean_feedback_forward_cosine": 0.9997610988155488, + "mean_feedback_forward_relative_error": 0.021491994029812275, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.999638557434082, + 0.9996102452278137, + 0.9996362924575806, + 0.9995453357696533, + 0.9996436834335327, + 0.9995719194412231, + 0.9996561408042908, + 0.999657154083252, + 0.999669075012207, + 0.9996402859687805, + 0.9996742010116577, + 0.9996456503868103, + 0.999691903591156, + 0.9995970129966736, + 0.9997073411941528, + 0.9996359348297119, + 0.9997234344482422, + 0.999672532081604, + 0.9997422695159912, + 0.9996891617774963, + 0.9997507333755493, + 0.9997844696044922, + 0.9997909069061279, + 0.9997697472572327, + 0.9998173713684082, + 0.9998141527175903, + 0.9998586177825928, + 0.9998947381973267, + 0.9999035596847534, + 0.9999581575393677, + 0.7821329832077026 + ], + "teaching_negative_gradient_cosine": [ + 0.999638557434082, + 0.9996102452278137, + 0.9996362924575806, + 0.9995453357696533, + 0.9996436834335327, + 0.9995719194412231, + 0.9996561408042908, + 0.999657154083252, + 0.999669075012207, + 0.9996402859687805, + 0.9996742010116577, + 0.9996456503868103, + 0.999691903591156, + 0.9995970129966736, + 0.9997073411941528, + 0.9996359348297119, + 0.9997234344482422, + 0.999672532081604, + 0.9997422695159912, + 0.9996891617774963, + 0.9997507333755493, + 0.9997844696044922, + 0.9997909069061279, + 0.9997697472572327, + 0.9998173713684082, + 0.9998141527175903, + 0.9998586177825928, + 0.9998947381973267, + 0.9999035596847534, + 0.9999581575393677, + 0.7821329832077026 + ], + "wall_s": 0.12763762474060059 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.332453727722168, + "mean_feedback_forward_cosine": 0.28569606547394105, + "mean_feedback_forward_relative_error": 1.1923096929827044, + "min_feedback_forward_cosine": 0.11065932363271713 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.8864616674423218 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2947325706481934, + "mean_feedback_forward_cosine": 0.37175222558359944, + "mean_feedback_forward_relative_error": 1.1189814344529183, + "min_feedback_forward_cosine": 0.16187171638011932 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.3447055550384521 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2630192041397095, + "mean_feedback_forward_cosine": 0.4468510333568819, + "mean_feedback_forward_relative_error": 1.050396325126771, + "min_feedback_forward_cosine": 0.20473386347293854 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0783762699890136 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2285550832748413, + "mean_feedback_forward_cosine": 0.506152376051872, + "mean_feedback_forward_relative_error": 0.9932994640642597, + "min_feedback_forward_cosine": 0.25025951862335205 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8908661921310425 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1947357654571533, + "mean_feedback_forward_cosine": 0.5549704615146883, + "mean_feedback_forward_relative_error": 0.9433515715983606, + "min_feedback_forward_cosine": 0.29240745306015015 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7713095247745514 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1657177209854126, + "mean_feedback_forward_cosine": 0.5965976099814138, + "mean_feedback_forward_relative_error": 0.898323812792378, + "min_feedback_forward_cosine": 0.333939790725708 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6912768111038208 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1398651599884033, + "mean_feedback_forward_cosine": 0.6339097003782949, + "mean_feedback_forward_relative_error": 0.8561380467107219, + "min_feedback_forward_cosine": 0.36467909812927246 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6317887006759644 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1156481504440308, + "mean_feedback_forward_cosine": 0.6650987209812287, + "mean_feedback_forward_relative_error": 0.8189005140335329, + "min_feedback_forward_cosine": 0.3936135768890381 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5794082787704468 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0911922454833984, + "mean_feedback_forward_cosine": 0.6936676425318564, + "mean_feedback_forward_relative_error": 0.7829820350293191, + "min_feedback_forward_cosine": 0.4210509657859802 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5458818233299255 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0665953159332275, + "mean_feedback_forward_cosine": 0.7185857988172962, + "mean_feedback_forward_relative_error": 0.7501856107865611, + "min_feedback_forward_cosine": 0.4483541250228882 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5144352616310119 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0408952236175537, + "mean_feedback_forward_cosine": 0.7419745210678347, + "mean_feedback_forward_relative_error": 0.7179888842567321, + "min_feedback_forward_cosine": 0.4757261276245117 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.48909983785629274 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0145304203033447, + "mean_feedback_forward_cosine": 0.7628409727927177, + "mean_feedback_forward_relative_error": 0.6880684966041196, + "min_feedback_forward_cosine": 0.5025681853294373 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.46696600249290465 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9898214936256409, + "mean_feedback_forward_cosine": 0.7811945292257494, + "mean_feedback_forward_relative_error": 0.660570744545229, + "min_feedback_forward_cosine": 0.5271745324134827 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.44200228766441346 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.965601921081543, + "mean_feedback_forward_cosine": 0.7980669429225307, + "mean_feedback_forward_relative_error": 0.6340958384736892, + "min_feedback_forward_cosine": 0.550247311592102 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.41981414313316345 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9398868083953857, + "mean_feedback_forward_cosine": 0.8137554430192516, + "mean_feedback_forward_relative_error": 0.608631880052628, + "min_feedback_forward_cosine": 0.5739184617996216 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.40447162753105165 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9141412973403931, + "mean_feedback_forward_cosine": 0.828351866814398, + "mean_feedback_forward_relative_error": 0.5839648977402718, + "min_feedback_forward_cosine": 0.597303032875061 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39567691297531127 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.889471173286438, + "mean_feedback_forward_cosine": 0.8413230834468719, + "mean_feedback_forward_relative_error": 0.5611338033791511, + "min_feedback_forward_cosine": 0.6190862655639648 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.37986853382110597 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8651845455169678, + "mean_feedback_forward_cosine": 0.8531636115043394, + "mean_feedback_forward_relative_error": 0.539392416515658, + "min_feedback_forward_cosine": 0.6397159099578857 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.36365509808063506 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.839250922203064, + "mean_feedback_forward_cosine": 0.8643700384324596, + "mean_feedback_forward_relative_error": 0.5179809389575836, + "min_feedback_forward_cosine": 0.6609064340591431 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3630038592338562 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8157009482383728, + "mean_feedback_forward_cosine": 0.8745473238729662, + "mean_feedback_forward_relative_error": 0.4977624238498749, + "min_feedback_forward_cosine": 0.679825484752655 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.34615828795433046 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7918817400932312, + "mean_feedback_forward_cosine": 0.8837778452903994, + "mean_feedback_forward_relative_error": 0.4788322655423995, + "min_feedback_forward_cosine": 0.6985763311386108 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3352479345989227 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7668979167938232, + "mean_feedback_forward_cosine": 0.8925440811341808, + "mean_feedback_forward_relative_error": 0.4600896590179013, + "min_feedback_forward_cosine": 0.7172738909721375 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3282865807914734 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7425411939620972, + "mean_feedback_forward_cosine": 0.900546173895559, + "mean_feedback_forward_relative_error": 0.4423999296080682, + "min_feedback_forward_cosine": 0.7347917556762695 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3158574280834198 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.719512939453125, + "mean_feedback_forward_cosine": 0.9078300268419327, + "mean_feedback_forward_relative_error": 0.42559024835786513, + "min_feedback_forward_cosine": 0.7508101463317871 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.30837035751342773 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6968992352485657, + "mean_feedback_forward_cosine": 0.9147653137483904, + "mean_feedback_forward_relative_error": 0.40906073008814164, + "min_feedback_forward_cosine": 0.766478955745697 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.29827840731620786 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6743670701980591, + "mean_feedback_forward_cosine": 0.9211131814987429, + "mean_feedback_forward_relative_error": 0.39326262089513964, + "min_feedback_forward_cosine": 0.781310498714447 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2950888151741028 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.651057243347168, + "mean_feedback_forward_cosine": 0.9269627544187731, + "mean_feedback_forward_relative_error": 0.37825878493247495, + "min_feedback_forward_cosine": 0.796197235584259 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2906055842828751 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6293326616287231, + "mean_feedback_forward_cosine": 0.9323475995371419, + "mean_feedback_forward_relative_error": 0.36379277369668406, + "min_feedback_forward_cosine": 0.8094980716705322 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2810540374565125 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6074268817901611, + "mean_feedback_forward_cosine": 0.9374242501874124, + "mean_feedback_forward_relative_error": 0.34971984788294763, + "min_feedback_forward_cosine": 0.8221464157104492 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2781663143348694 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5866720080375671, + "mean_feedback_forward_cosine": 0.9419786180219343, + "mean_feedback_forward_relative_error": 0.33660862138194425, + "min_feedback_forward_cosine": 0.8340702056884766 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.27273870217323304 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5648839473724365, + "mean_feedback_forward_cosine": 0.9463978294403322, + "mean_feedback_forward_relative_error": 0.3234559158163686, + "min_feedback_forward_cosine": 0.8459969162940979 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26634878540039064 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5445200800895691, + "mean_feedback_forward_cosine": 0.9504142595875648, + "mean_feedback_forward_relative_error": 0.31098151182936085, + "min_feedback_forward_cosine": 0.856764554977417 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26126067794322966 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5249777436256409, + "mean_feedback_forward_cosine": 0.9541119606264176, + "mean_feedback_forward_relative_error": 0.29899491441826664, + "min_feedback_forward_cosine": 0.8667631149291992 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2520839551353455 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5054250359535217, + "mean_feedback_forward_cosine": 0.9575571475490448, + "mean_feedback_forward_relative_error": 0.28742779647150346, + "min_feedback_forward_cosine": 0.8762861490249634 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25377710359573363 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48607558012008667, + "mean_feedback_forward_cosine": 0.9607740486821821, + "mean_feedback_forward_relative_error": 0.2762442082166672, + "min_feedback_forward_cosine": 0.8854256868362427 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.24896999197483063 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4677378535270691, + "mean_feedback_forward_cosine": 0.963708339198943, + "mean_feedback_forward_relative_error": 0.26564024220551213, + "min_feedback_forward_cosine": 0.8937792181968689 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2468858616542816 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4498147666454315, + "mean_feedback_forward_cosine": 0.9664514295516475, + "mean_feedback_forward_relative_error": 0.2553777086638635, + "min_feedback_forward_cosine": 0.9017024040222168 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.23782768265247345 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43192774057388306, + "mean_feedback_forward_cosine": 0.9689808353301017, + "mean_feedback_forward_relative_error": 0.24550913610766012, + "min_feedback_forward_cosine": 0.9092162847518921 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.23959526492118835 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41520217061042786, + "mean_feedback_forward_cosine": 0.9713096599425038, + "mean_feedback_forward_relative_error": 0.23603895787269838, + "min_feedback_forward_cosine": 0.9160403609275818 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23518191647052764 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.39894381165504456, + "mean_feedback_forward_cosine": 0.9734609646181906, + "mean_feedback_forward_relative_error": 0.2269706735687871, + "min_feedback_forward_cosine": 0.9224241971969604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23041978515148162 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38303983211517334, + "mean_feedback_forward_cosine": 0.975447416305542, + "mean_feedback_forward_relative_error": 0.2182956261019553, + "min_feedback_forward_cosine": 0.92841637134552 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23074689755916594 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.368954598903656, + "mean_feedback_forward_cosine": 0.9772347577156559, + "mean_feedback_forward_relative_error": 0.21015597807784234, + "min_feedback_forward_cosine": 0.9335581064224243 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22321409762859346 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3541412353515625, + "mean_feedback_forward_cosine": 0.9790231431684187, + "mean_feedback_forward_relative_error": 0.2016921742789207, + "min_feedback_forward_cosine": 0.9387012124061584 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22403297308921813 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33939701318740845, + "mean_feedback_forward_cosine": 0.9805971268684633, + "mean_feedback_forward_relative_error": 0.19396936436814646, + "min_feedback_forward_cosine": 0.9436639547348022 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22272661297798158 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32659563422203064, + "mean_feedback_forward_cosine": 0.98200483475962, + "mean_feedback_forward_relative_error": 0.18678255018688017, + "min_feedback_forward_cosine": 0.9477999210357666 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21533804458618164 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3134482502937317, + "mean_feedback_forward_cosine": 0.9833629381272101, + "mean_feedback_forward_relative_error": 0.17959226283334917, + "min_feedback_forward_cosine": 0.9518828392028809 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21674478680610657 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3012486398220062, + "mean_feedback_forward_cosine": 0.9845931433862255, + "mean_feedback_forward_relative_error": 0.17279463826167968, + "min_feedback_forward_cosine": 0.9555051326751709 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2155820286989212 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28929030895233154, + "mean_feedback_forward_cosine": 0.985737721766195, + "mean_feedback_forward_relative_error": 0.1662241930682813, + "min_feedback_forward_cosine": 0.9589306116104126 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2111910402202606 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2772831618785858, + "mean_feedback_forward_cosine": 0.986845201061618, + "mean_feedback_forward_relative_error": 0.15963495522737503, + "min_feedback_forward_cosine": 0.9622498154640198 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21108394720077514 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26609328389167786, + "mean_feedback_forward_cosine": 0.9878137342391475, + "mean_feedback_forward_relative_error": 0.1536394064464877, + "min_feedback_forward_cosine": 0.9651902914047241 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20221650484085082 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25522199273109436, + "mean_feedback_forward_cosine": 0.9887455471100346, + "mean_feedback_forward_relative_error": 0.1476403996348381, + "min_feedback_forward_cosine": 0.9679491519927979 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20754753598213196 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2448800802230835, + "mean_feedback_forward_cosine": 0.9895845670853892, + "mean_feedback_forward_relative_error": 0.14203302297861345, + "min_feedback_forward_cosine": 0.9704720973968506 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.2019150578212738 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23494644463062286, + "mean_feedback_forward_cosine": 0.9903780145029868, + "mean_feedback_forward_relative_error": 0.13650183896383933, + "min_feedback_forward_cosine": 0.9727993607521057 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20128608113765717 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22498035430908203, + "mean_feedback_forward_cosine": 0.9911072619499699, + "mean_feedback_forward_relative_error": 0.1312497033467216, + "min_feedback_forward_cosine": 0.9750555157661438 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.20113794761180878 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21596264839172363, + "mean_feedback_forward_cosine": 0.9917765778879966, + "mean_feedback_forward_relative_error": 0.1262037168827749, + "min_feedback_forward_cosine": 0.9769986867904663 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19641307671546937 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20744924247264862, + "mean_feedback_forward_cosine": 0.9923929975878808, + "mean_feedback_forward_relative_error": 0.12137739264195965, + "min_feedback_forward_cosine": 0.9787628054618835 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1954195665884018 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19935357570648193, + "mean_feedback_forward_cosine": 0.9929594359090251, + "mean_feedback_forward_relative_error": 0.11676642827449306, + "min_feedback_forward_cosine": 0.9803847670555115 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19436430928230286 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19114062190055847, + "mean_feedback_forward_cosine": 0.9934934050806107, + "mean_feedback_forward_relative_error": 0.11225502865929757, + "min_feedback_forward_cosine": 0.9819468259811401 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19422180081844329 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.183742955327034, + "mean_feedback_forward_cosine": 0.9939748471783053, + "mean_feedback_forward_relative_error": 0.10800776967117863, + "min_feedback_forward_cosine": 0.983309268951416 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.1890528978586197 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17630770802497864, + "mean_feedback_forward_cosine": 0.9944276944283517, + "mean_feedback_forward_relative_error": 0.1038708767342952, + "min_feedback_forward_cosine": 0.9846312999725342 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1900973216152191 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16908420622348785, + "mean_feedback_forward_cosine": 0.9948449615509279, + "mean_feedback_forward_relative_error": 0.09990192457072196, + "min_feedback_forward_cosine": 0.985857367515564 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19067359004735945 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16221991181373596, + "mean_feedback_forward_cosine": 0.9952308247166295, + "mean_feedback_forward_relative_error": 0.09608427531296207, + "min_feedback_forward_cosine": 0.9869732856750488 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18816538153648377 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1557830423116684, + "mean_feedback_forward_cosine": 0.9955845071423438, + "mean_feedback_forward_relative_error": 0.0924553124895019, + "min_feedback_forward_cosine": 0.9879834651947021 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1851732423210144 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14936737716197968, + "mean_feedback_forward_cosine": 0.9959167568914352, + "mean_feedback_forward_relative_error": 0.0889079782270616, + "min_feedback_forward_cosine": 0.9889500141143799 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18822476876735686 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14342321455478668, + "mean_feedback_forward_cosine": 0.9962218461498138, + "mean_feedback_forward_relative_error": 0.08552141439530157, + "min_feedback_forward_cosine": 0.989808201789856 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18450284621238708 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13741940259933472, + "mean_feedback_forward_cosine": 0.9965088386689464, + "mean_feedback_forward_relative_error": 0.0822157223498629, + "min_feedback_forward_cosine": 0.9906416535377502 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1846496206665039 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1318826675415039, + "mean_feedback_forward_cosine": 0.9967711721697161, + "mean_feedback_forward_relative_error": 0.07906677039159883, + "min_feedback_forward_cosine": 0.991375207901001 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18396508299827577 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12647992372512817, + "mean_feedback_forward_cosine": 0.9970187756323046, + "mean_feedback_forward_relative_error": 0.07597769808865362, + "min_feedback_forward_cosine": 0.9920645952224731 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18446759902000429 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12145280092954636, + "mean_feedback_forward_cosine": 0.9972406164292367, + "mean_feedback_forward_relative_error": 0.07309857100969361, + "min_feedback_forward_cosine": 0.992681086063385 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18181212815999984 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11659641563892365, + "mean_feedback_forward_cosine": 0.9974452833975515, + "mean_feedback_forward_relative_error": 0.07033463843887852, + "min_feedback_forward_cosine": 0.9932525157928467 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17683307625293732 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1117180585861206, + "mean_feedback_forward_cosine": 0.9976399214037003, + "mean_feedback_forward_relative_error": 0.06761032252782775, + "min_feedback_forward_cosine": 0.993803858757019 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1790683585214615 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10718255490064621, + "mean_feedback_forward_cosine": 0.9978178970275386, + "mean_feedback_forward_relative_error": 0.06500570896652437, + "min_feedback_forward_cosine": 0.9942947626113892 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1744948213863373 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10274895280599594, + "mean_feedback_forward_cosine": 0.9979846592872373, + "mean_feedback_forward_relative_error": 0.062479867930373835, + "min_feedback_forward_cosine": 0.9947553873062134 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18304942828178405 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09887866675853729, + "mean_feedback_forward_cosine": 0.9981283007129547, + "mean_feedback_forward_relative_error": 0.06020873045969394, + "min_feedback_forward_cosine": 0.9951422810554504 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17089463544845582 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09468246251344681, + "mean_feedback_forward_cosine": 0.9982725804851901, + "mean_feedback_forward_relative_error": 0.05784883536398411, + "min_feedback_forward_cosine": 0.9955440759658813 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17783002980709076 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09086678177118301, + "mean_feedback_forward_cosine": 0.9984012207677287, + "mean_feedback_forward_relative_error": 0.05565292400217826, + "min_feedback_forward_cosine": 0.995894193649292 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1737877887916565 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08716202527284622, + "mean_feedback_forward_cosine": 0.998520556957491, + "mean_feedback_forward_relative_error": 0.05353486916470912, + "min_feedback_forward_cosine": 0.9962215423583984 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17157120014190674 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08369570225477219, + "mean_feedback_forward_cosine": 0.9986321349297801, + "mean_feedback_forward_relative_error": 0.05147713903457888, + "min_feedback_forward_cosine": 0.9965160489082336 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17163206961631775 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0802949070930481, + "mean_feedback_forward_cosine": 0.9987352817289291, + "mean_feedback_forward_relative_error": 0.0495004809551662, + "min_feedback_forward_cosine": 0.9967924952507019 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1683024906373024 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07682712376117706, + "mean_feedback_forward_cosine": 0.9988350791315879, + "mean_feedback_forward_relative_error": 0.0475140098602541, + "min_feedback_forward_cosine": 0.9970628023147583 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1741444329929352 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07371344417333603, + "mean_feedback_forward_cosine": 0.9989226691184505, + "mean_feedback_forward_relative_error": 0.04569508850334152, + "min_feedback_forward_cosine": 0.9972962141036987 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1721878899717331 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07070252299308777, + "mean_feedback_forward_cosine": 0.9990031642298545, + "mean_feedback_forward_relative_error": 0.0439572018841582, + "min_feedback_forward_cosine": 0.9975121021270752 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16856893158912659 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06776734441518784, + "mean_feedback_forward_cosine": 0.9990783141505334, + "mean_feedback_forward_relative_error": 0.04227025079871378, + "min_feedback_forward_cosine": 0.9977139234542847 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16814912214279173 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06500975042581558, + "mean_feedback_forward_cosine": 0.9991486053312978, + "mean_feedback_forward_relative_error": 0.04062697895231747, + "min_feedback_forward_cosine": 0.9978955984115601 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.16887235580444335 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06233934313058853, + "mean_feedback_forward_cosine": 0.9992128937475143, + "mean_feedback_forward_relative_error": 0.039068897134594376, + "min_feedback_forward_cosine": 0.9980646371841431 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16984825755119323 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05984418839216232, + "mean_feedback_forward_cosine": 0.9992719600277562, + "mean_feedback_forward_relative_error": 0.037575110522729736, + "min_feedback_forward_cosine": 0.9982162714004517 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16876170399665832 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0574023611843586, + "mean_feedback_forward_cosine": 0.9993275230930697, + "mean_feedback_forward_relative_error": 0.036111079306612094, + "min_feedback_forward_cosine": 0.9983586072921753 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16502491756916046 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05504690110683441, + "mean_feedback_forward_cosine": 0.9993792676156567, + "mean_feedback_forward_relative_error": 0.034695310157633594, + "min_feedback_forward_cosine": 0.9984903335571289 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.1693146181845665 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.052905917167663574, + "mean_feedback_forward_cosine": 0.9994260822573016, + "mean_feedback_forward_relative_error": 0.03336298264442913, + "min_feedback_forward_cosine": 0.998605489730835 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16608294499397278 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05085453763604164, + "mean_feedback_forward_cosine": 0.9994678977997072, + "mean_feedback_forward_relative_error": 0.032121037133038044, + "min_feedback_forward_cosine": 0.9987109899520874 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16007911418914794 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04878352954983711, + "mean_feedback_forward_cosine": 0.9995081136303563, + "mean_feedback_forward_relative_error": 0.030884851701557636, + "min_feedback_forward_cosine": 0.9988136887550354 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1665627337384224 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04683680459856987, + "mean_feedback_forward_cosine": 0.9995447685641627, + "mean_feedback_forward_relative_error": 0.029712052264761542, + "min_feedback_forward_cosine": 0.9989064335823059 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16070930757522583 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.044917646795511246, + "mean_feedback_forward_cosine": 0.9995796122858601, + "mean_feedback_forward_relative_error": 0.028554636534423597, + "min_feedback_forward_cosine": 0.9989941120147705 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1618361934566498 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.043176449835300446, + "mean_feedback_forward_cosine": 0.9996113450296463, + "mean_feedback_forward_relative_error": 0.027455753827046965, + "min_feedback_forward_cosine": 0.999070405960083 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16187013481616974 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.041375935077667236, + "mean_feedback_forward_cosine": 0.9996412434885579, + "mean_feedback_forward_relative_error": 0.02638026164664376, + "min_feedback_forward_cosine": 0.9991462826728821 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1638966773366928 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039716508239507675, + "mean_feedback_forward_cosine": 0.9996683982110792, + "mean_feedback_forward_relative_error": 0.02536310447800544, + "min_feedback_forward_cosine": 0.9992133378982544 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16252423438310623 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038125619292259216, + "mean_feedback_forward_cosine": 0.9996935590620963, + "mean_feedback_forward_relative_error": 0.024381615101329742, + "min_feedback_forward_cosine": 0.9992749691009521 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.16267103942394256 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.036539141088724136, + "mean_feedback_forward_cosine": 0.9997172394106465, + "mean_feedback_forward_relative_error": 0.023422856334476702, + "min_feedback_forward_cosine": 0.9993340969085693 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16701118450164795 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03513658419251442, + "mean_feedback_forward_cosine": 0.9997374261579206, + "mean_feedback_forward_relative_error": 0.022570521690912785, + "min_feedback_forward_cosine": 0.9993841648101807 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.15273886563301087 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03375523164868355, + "mean_feedback_forward_cosine": 0.9997568380448126, + "mean_feedback_forward_relative_error": 0.021721016255117232, + "min_feedback_forward_cosine": 0.9994314908981323 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15564070889234544 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033744946122169495, + "mean_feedback_forward_cosine": 0.9997570226269383, + "mean_feedback_forward_relative_error": 0.021712409422522592, + "min_feedback_forward_cosine": 0.9994319677352905 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08587176040172577 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03373920917510986, + "mean_feedback_forward_cosine": 0.9997570918452355, + "mean_feedback_forward_relative_error": 0.02170650997469502, + "min_feedback_forward_cosine": 0.9994320869445801 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.064969905834198 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0337340384721756, + "mean_feedback_forward_cosine": 0.9997572072090641, + "mean_feedback_forward_relative_error": 0.021701117451753345, + "min_feedback_forward_cosine": 0.9994323253631592 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.054667386298179625 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03372880443930626, + "mean_feedback_forward_cosine": 0.9997573129592403, + "mean_feedback_forward_relative_error": 0.021695784263072476, + "min_feedback_forward_cosine": 0.9994325637817383 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04887079746007919 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0337241068482399, + "mean_feedback_forward_cosine": 0.9997573667956937, + "mean_feedback_forward_relative_error": 0.021690729569884076, + "min_feedback_forward_cosine": 0.9994326233863831 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.0453762242937088 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03371940553188324, + "mean_feedback_forward_cosine": 0.9997574360139908, + "mean_feedback_forward_relative_error": 0.021685759657092633, + "min_feedback_forward_cosine": 0.9994326829910278 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.041495048912763594 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033714789897203445, + "mean_feedback_forward_cosine": 0.9997575321505147, + "mean_feedback_forward_relative_error": 0.02168095063778662, + "min_feedback_forward_cosine": 0.9994329214096069 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.037941184784173965 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033710163086652756, + "mean_feedback_forward_cosine": 0.9997576052142728, + "mean_feedback_forward_relative_error": 0.021676265259063052, + "min_feedback_forward_cosine": 0.9994330406188965 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03754971484422684 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03370567783713341, + "mean_feedback_forward_cosine": 0.9997577013507966, + "mean_feedback_forward_relative_error": 0.02167169302101097, + "min_feedback_forward_cosine": 0.9994332194328308 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03408987087845802 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03370092064142227, + "mean_feedback_forward_cosine": 0.9997577782600157, + "mean_feedback_forward_relative_error": 0.021667062049551357, + "min_feedback_forward_cosine": 0.9994333386421204 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.032719900552630424 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033696338534355164, + "mean_feedback_forward_cosine": 0.9997578570919652, + "mean_feedback_forward_relative_error": 0.0216625030302713, + "min_feedback_forward_cosine": 0.9994335174560547 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.031641623229980466 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033692020922899246, + "mean_feedback_forward_cosine": 0.9997579416921062, + "mean_feedback_forward_relative_error": 0.021658158695866025, + "min_feedback_forward_cosine": 0.999433696269989 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.028986293906569482 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03368794173002243, + "mean_feedback_forward_cosine": 0.9997580224467862, + "mean_feedback_forward_relative_error": 0.021653961570512865, + "min_feedback_forward_cosine": 0.9994337558746338 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.026132644518613815 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033683449029922485, + "mean_feedback_forward_cosine": 0.9997580955105443, + "mean_feedback_forward_relative_error": 0.021649496299364874, + "min_feedback_forward_cosine": 0.9994339346885681 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02686688132405281 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03367915004491806, + "mean_feedback_forward_cosine": 0.9997581685743024, + "mean_feedback_forward_relative_error": 0.02164519934224025, + "min_feedback_forward_cosine": 0.9994341731071472 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.025645499723553657 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03367479518055916, + "mean_feedback_forward_cosine": 0.9997582397153301, + "mean_feedback_forward_relative_error": 0.02164089640662555, + "min_feedback_forward_cosine": 0.9994341731071472 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.02355453469067812 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033670708537101746, + "mean_feedback_forward_cosine": 0.9997583147018186, + "mean_feedback_forward_relative_error": 0.021636745396761164, + "min_feedback_forward_cosine": 0.9994344115257263 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.02261603930771351 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0336662121117115, + "mean_feedback_forward_cosine": 0.999758405070151, + "mean_feedback_forward_relative_error": 0.021632374309363864, + "min_feedback_forward_cosine": 0.9994345903396606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02236939556080848 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033661872148513794, + "mean_feedback_forward_cosine": 0.9997584762111786, + "mean_feedback_forward_relative_error": 0.02162814729156033, + "min_feedback_forward_cosine": 0.9994347095489502 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.021888015147149564 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03365737572312355, + "mean_feedback_forward_cosine": 0.9997585531203977, + "mean_feedback_forward_relative_error": 0.02162377931357872, + "min_feedback_forward_cosine": 0.9994348287582397 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.021143408103883266 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033653125166893005, + "mean_feedback_forward_cosine": 0.9997586531023825, + "mean_feedback_forward_relative_error": 0.021619573505895752, + "min_feedback_forward_cosine": 0.9994350075721741 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.020178167527318002 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0336487852036953, + "mean_feedback_forward_cosine": 0.9997587223206797, + "mean_feedback_forward_relative_error": 0.021615374818324082, + "min_feedback_forward_cosine": 0.9994350671768188 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.019790309482812882 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033644478768110275, + "mean_feedback_forward_cosine": 0.9997587819253245, + "mean_feedback_forward_relative_error": 0.021611214915831244, + "min_feedback_forward_cosine": 0.9994351863861084 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.01804049372404814 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033640433102846146, + "mean_feedback_forward_cosine": 0.9997588569118131, + "mean_feedback_forward_relative_error": 0.021607064461756136, + "min_feedback_forward_cosine": 0.999435305595398 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.017340303631424903 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03363589942455292, + "mean_feedback_forward_cosine": 0.9997589549710674, + "mean_feedback_forward_relative_error": 0.021602735839663975, + "min_feedback_forward_cosine": 0.9994354248046875 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01802509668469429 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03363114595413208, + "mean_feedback_forward_cosine": 0.9997590453393997, + "mean_feedback_forward_relative_error": 0.021598194245128863, + "min_feedback_forward_cosine": 0.9994357228279114 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.018359613683819772 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03362688049674034, + "mean_feedback_forward_cosine": 0.9997591472441151, + "mean_feedback_forward_relative_error": 0.02159395014807101, + "min_feedback_forward_cosine": 0.9994357824325562 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.01639207499742508 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03362235799431801, + "mean_feedback_forward_cosine": 0.9997592241533341, + "mean_feedback_forward_relative_error": 0.02158959495324281, + "min_feedback_forward_cosine": 0.9994359612464905 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01609149567514658 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03361808881163597, + "mean_feedback_forward_cosine": 0.9997592914489007, + "mean_feedback_forward_relative_error": 0.02158542512164962, + "min_feedback_forward_cosine": 0.9994361400604248 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.016263673985898494 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033613912761211395, + "mean_feedback_forward_cosine": 0.9997593664353893, + "mean_feedback_forward_relative_error": 0.021581318950460802, + "min_feedback_forward_cosine": 0.9994363188743591 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.014643422370404006 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033610012382268906, + "mean_feedback_forward_cosine": 0.9997594471900694, + "mean_feedback_forward_relative_error": 0.0215773728616055, + "min_feedback_forward_cosine": 0.9994363784790039 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.014087463182508945 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03360601142048836, + "mean_feedback_forward_cosine": 0.9997595356356713, + "mean_feedback_forward_relative_error": 0.021573296342525753, + "min_feedback_forward_cosine": 0.9994365572929382 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.013209543904662132 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033601053059101105, + "mean_feedback_forward_cosine": 0.9997596106221599, + "mean_feedback_forward_relative_error": 0.021568678244347533, + "min_feedback_forward_cosine": 0.9994366765022278 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.014964622594416142 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03359664976596832, + "mean_feedback_forward_cosine": 0.999759687531379, + "mean_feedback_forward_relative_error": 0.021564522322507634, + "min_feedback_forward_cosine": 0.9994367361068726 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.013909739240705966 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03359221667051315, + "mean_feedback_forward_cosine": 0.9997597875133637, + "mean_feedback_forward_relative_error": 0.021560262798541976, + "min_feedback_forward_cosine": 0.9994370341300964 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013652800155878067 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03358719125390053, + "mean_feedback_forward_cosine": 0.9997598586543914, + "mean_feedback_forward_relative_error": 0.021555697725665186, + "min_feedback_forward_cosine": 0.999437153339386 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.014037335795909166 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033582653850317, + "mean_feedback_forward_cosine": 0.9997599624818371, + "mean_feedback_forward_relative_error": 0.021551487862222617, + "min_feedback_forward_cosine": 0.9994373917579651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.013184962499141693 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03357814997434616, + "mean_feedback_forward_cosine": 0.9997600374683258, + "mean_feedback_forward_relative_error": 0.021547226896209103, + "min_feedback_forward_cosine": 0.9994373917579651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.0119105725453794 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033573850989341736, + "mean_feedback_forward_cosine": 0.9997601336048495, + "mean_feedback_forward_relative_error": 0.021543031573415763, + "min_feedback_forward_cosine": 0.9994375705718994 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012246701563745736 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03356962278485298, + "mean_feedback_forward_cosine": 0.9997602085913381, + "mean_feedback_forward_relative_error": 0.021538912513924223, + "min_feedback_forward_cosine": 0.9994378685951233 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012045851690918208 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03356557339429855, + "mean_feedback_forward_cosine": 0.9997602720414439, + "mean_feedback_forward_relative_error": 0.02153489657587582, + "min_feedback_forward_cosine": 0.9994379281997681 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.011227504568099976 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03356126323342323, + "mean_feedback_forward_cosine": 0.99976033741428, + "mean_feedback_forward_relative_error": 0.02153074730848593, + "min_feedback_forward_cosine": 0.9994380474090576 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011131018073707819 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03355677053332329, + "mean_feedback_forward_cosine": 0.9997604354735343, + "mean_feedback_forward_relative_error": 0.02152619996078072, + "min_feedback_forward_cosine": 0.9994382262229919 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011638076515942812 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033552322536706924, + "mean_feedback_forward_cosine": 0.9997605604510154, + "mean_feedback_forward_relative_error": 0.0215218830553274, + "min_feedback_forward_cosine": 0.9994382858276367 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.010579599971473216 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03354793041944504, + "mean_feedback_forward_cosine": 0.9997606412056954, + "mean_feedback_forward_relative_error": 0.021517586593906724, + "min_feedback_forward_cosine": 0.999438464641571 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.01081768324844539 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03354306146502495, + "mean_feedback_forward_cosine": 0.9997607065785316, + "mean_feedback_forward_relative_error": 0.02151310071349144, + "min_feedback_forward_cosine": 0.9994386434555054 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010473076796680688 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03353889286518097, + "mean_feedback_forward_cosine": 0.9997607854104811, + "mean_feedback_forward_relative_error": 0.021509090183122505, + "min_feedback_forward_cosine": 0.9994387626647949 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.009850473795235157 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03353442624211311, + "mean_feedback_forward_cosine": 0.999760877701544, + "mean_feedback_forward_relative_error": 0.021504730932534702, + "min_feedback_forward_cosine": 0.9994389414787292 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.010320796567499638 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352973610162735, + "mean_feedback_forward_cosine": 0.9997609488425716, + "mean_feedback_forward_relative_error": 0.021500151529307326, + "min_feedback_forward_cosine": 0.9994391202926636 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.01045252704411745 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0335252583026886, + "mean_feedback_forward_cosine": 0.9997610526700174, + "mean_feedback_forward_relative_error": 0.021495793149956772, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009837141685783863 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352520614862442, + "mean_feedback_forward_cosine": 0.9997610411336345, + "mean_feedback_forward_relative_error": 0.02149569653275032, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.009407439221739769 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352515771985054, + "mean_feedback_forward_cosine": 0.9997610507472869, + "mean_feedback_forward_relative_error": 0.021495603745983492, + "min_feedback_forward_cosine": 0.9994391202926636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.009692396873384714 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352513164281845, + "mean_feedback_forward_cosine": 0.9997610776655136, + "mean_feedback_forward_relative_error": 0.02149552321662345, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.00818301939830184 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033525098115205765, + "mean_feedback_forward_cosine": 0.9997610372881736, + "mean_feedback_forward_relative_error": 0.021495435477024125, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.007863032380789519 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033525072038173676, + "mean_feedback_forward_cosine": 0.9997610680518612, + "mean_feedback_forward_relative_error": 0.02149535407642684, + "min_feedback_forward_cosine": 0.9994391202926636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.007916177957579493 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352503851056099, + "mean_feedback_forward_cosine": 0.9997610680518612, + "mean_feedback_forward_relative_error": 0.021495273757365443, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007483635274618864 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0335250049829483, + "mean_feedback_forward_cosine": 0.9997610584382088, + "mean_feedback_forward_relative_error": 0.02149519364860269, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007406631565839052 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352498635649681, + "mean_feedback_forward_cosine": 0.9997610526700174, + "mean_feedback_forward_relative_error": 0.02149511546257042, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.006852073890566826 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352494537830353, + "mean_feedback_forward_cosine": 0.999761079588244, + "mean_feedback_forward_relative_error": 0.021495032935373245, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007527244059741497 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352490812540054, + "mean_feedback_forward_cosine": 0.9997610565154783, + "mean_feedback_forward_relative_error": 0.02149495154979729, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.00713704448774457 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524878323078156, + "mean_feedback_forward_cosine": 0.9997610680518612, + "mean_feedback_forward_relative_error": 0.021494870119157336, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0073489983118325474 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352484479546547, + "mean_feedback_forward_cosine": 0.9997610815109745, + "mean_feedback_forward_relative_error": 0.021494792654148994, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.006711921958327293 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352481499314308, + "mean_feedback_forward_cosine": 0.9997610661291307, + "mean_feedback_forward_relative_error": 0.0214947137020288, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.006963134993463755 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352478891611099, + "mean_feedback_forward_cosine": 0.9997610776655136, + "mean_feedback_forward_relative_error": 0.02149463832498558, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.006833728794082999 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524755388498306, + "mean_feedback_forward_cosine": 0.9997610622836698, + "mean_feedback_forward_relative_error": 0.02149456064967859, + "min_feedback_forward_cosine": 0.9994393587112427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0069753591714054345 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352471813559532, + "mean_feedback_forward_cosine": 0.9997610872791659, + "mean_feedback_forward_relative_error": 0.02149448093147047, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.006809330035448075 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352469205856323, + "mean_feedback_forward_cosine": 0.9997610642064002, + "mean_feedback_forward_relative_error": 0.021494409384866877, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.006547005600109696 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524662256240845, + "mean_feedback_forward_cosine": 0.9997610776655136, + "mean_feedback_forward_relative_error": 0.021494334443442283, + "min_feedback_forward_cosine": 0.9994392991065979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006588688519597053 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352462872862816, + "mean_feedback_forward_cosine": 0.9997610603609393, + "mean_feedback_forward_relative_error": 0.021494263437606635, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0064581088467687365 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352459892630577, + "mean_feedback_forward_cosine": 0.9997610776655136, + "mean_feedback_forward_relative_error": 0.021494189517632607, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006974443473666906 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352456912398338, + "mean_feedback_forward_cosine": 0.9997610815109745, + "mean_feedback_forward_relative_error": 0.02149411502684797, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.006474740914031863 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524543046951294, + "mean_feedback_forward_cosine": 0.999761079588244, + "mean_feedback_forward_relative_error": 0.021494041302151257, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.005849080839082599 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0335245244204998, + "mean_feedback_forward_cosine": 0.9997610872791659, + "mean_feedback_forward_relative_error": 0.021493971332787506, + "min_feedback_forward_cosine": 0.9994393587112427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.005923768717944622 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352448716759682, + "mean_feedback_forward_cosine": 0.9997610680518612, + "mean_feedback_forward_relative_error": 0.02149389556518966, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006921178120821715 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352445363998413, + "mean_feedback_forward_cosine": 0.9997610699745917, + "mean_feedback_forward_relative_error": 0.02149382048857308, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006746948901340365 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352442756295204, + "mean_feedback_forward_cosine": 0.9997610757427831, + "mean_feedback_forward_relative_error": 0.02149374701923901, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006347868871837854 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524394035339355, + "mean_feedback_forward_cosine": 0.999761079588244, + "mean_feedback_forward_relative_error": 0.02149367281385968, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006598144030719995 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352436423301697, + "mean_feedback_forward_cosine": 0.9997610815109745, + "mean_feedback_forward_relative_error": 0.021493595514085986, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0065184420828521255 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352434188127518, + "mean_feedback_forward_cosine": 0.9997610853564355, + "mean_feedback_forward_relative_error": 0.021493520813002702, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.005984624702259898 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352431207895279, + "mean_feedback_forward_cosine": 0.9997610988155488, + "mean_feedback_forward_relative_error": 0.02149344841018319, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006538475824892521 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0335242785513401, + "mean_feedback_forward_cosine": 0.9997610930473574, + "mean_feedback_forward_relative_error": 0.02149337551165973, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0066132744911313055 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524252474308014, + "mean_feedback_forward_cosine": 0.9997610949700878, + "mean_feedback_forward_relative_error": 0.0214933046560374, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006083347657397389 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524226397275925, + "mean_feedback_forward_cosine": 0.9997610988155488, + "mean_feedback_forward_relative_error": 0.021493231066532673, + "min_feedback_forward_cosine": 0.9994392991065979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006523385307714343 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352418914437294, + "mean_feedback_forward_cosine": 0.999761083433705, + "mean_feedback_forward_relative_error": 0.02149315140840988, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006543258281052113 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352416306734085, + "mean_feedback_forward_cosine": 0.9997610872791659, + "mean_feedback_forward_relative_error": 0.021493081318875475, + "min_feedback_forward_cosine": 0.9994392991065979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.005880762688852847 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352414444088936, + "mean_feedback_forward_cosine": 0.9997610853564355, + "mean_feedback_forward_relative_error": 0.021493008180010702, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006099545330628752 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352411836385727, + "mean_feedback_forward_cosine": 0.9997610949700878, + "mean_feedback_forward_relative_error": 0.021492930985386332, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006418848850876093 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352409228682518, + "mean_feedback_forward_cosine": 0.9997610968928183, + "mean_feedback_forward_relative_error": 0.021492855668428443, + "min_feedback_forward_cosine": 0.9994393587112427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0057901674166321755 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352406248450279, + "mean_feedback_forward_cosine": 0.9997610872791659, + "mean_feedback_forward_relative_error": 0.021492780261342565, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0058912814247608185 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033524032682180405, + "mean_feedback_forward_cosine": 0.9997611065064708, + "mean_feedback_forward_relative_error": 0.021492710802704096, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006571254019960761 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352399542927742, + "mean_feedback_forward_cosine": 0.9997610930473574, + "mean_feedback_forward_relative_error": 0.021492637423498016, + "min_feedback_forward_cosine": 0.9994391798973083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006014063582122326 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352397680282593, + "mean_feedback_forward_cosine": 0.9997610892018964, + "mean_feedback_forward_relative_error": 0.021492564209526586, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006366922727227211 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352394700050354, + "mean_feedback_forward_cosine": 0.9997610872791659, + "mean_feedback_forward_relative_error": 0.021492496373191956, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006320757586359978 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352392092347145, + "mean_feedback_forward_cosine": 0.9997610930473574, + "mean_feedback_forward_relative_error": 0.021492425412420305, + "min_feedback_forward_cosine": 0.9994392991065979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005957902250476182 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352389484643936, + "mean_feedback_forward_cosine": 0.9997611045837402, + "mean_feedback_forward_relative_error": 0.02149235218342754, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006235929216071964 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352387621998787, + "mean_feedback_forward_cosine": 0.9997610911246269, + "mean_feedback_forward_relative_error": 0.021492283460834333, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005720531714558602 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352384641766548, + "mean_feedback_forward_cosine": 0.9997610892018964, + "mean_feedback_forward_relative_error": 0.021492208384217754, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0063629729913175104 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352382034063339, + "mean_feedback_forward_cosine": 0.9997611045837402, + "mean_feedback_forward_relative_error": 0.021492136056504903, + "min_feedback_forward_cosine": 0.9994392991065979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0061088988988101485 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033523790538311005, + "mean_feedback_forward_cosine": 0.9997610930473574, + "mean_feedback_forward_relative_error": 0.021492065621479865, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.006140658489614725 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352376073598862, + "mean_feedback_forward_cosine": 0.9997610988155488, + "mean_feedback_forward_relative_error": 0.021491994029812275, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.006112837376818061 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9258, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3920034271240234, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2525847040, + "peak_memory_reserved_bytes": 3393191936, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1602163314819336, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10196.078639745712, + "train_wall_s": 10192.868887662888 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684202240000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2745656320000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json new file mode 100644 index 0000000..96df080 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json @@ -0,0 +1,3654 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9994567665788863, + "feedback_forward_cosine": [ + 0.9995390176773071, + 0.9994896054267883, + 0.9993683695793152, + 0.9992607831954956, + 0.9991724491119385, + 0.9991496801376343, + 0.9994419813156128, + 0.9994545578956604, + 0.9993976354598999, + 0.9993428587913513, + 0.9996061325073242, + 0.9995467662811279, + 0.9996442794799805, + 0.9996175765991211, + 0.9997511506080627, + 0.9997085332870483, + 0.9997986555099487, + 0.9996837973594666, + 0.9996693134307861, + 0.9997653961181641, + 0.9995806813240051, + 0.99955153465271, + 0.9996522665023804, + 0.999573826789856, + 0.9996734857559204, + 0.9995680451393127, + 0.9996894598007202, + 0.9995835423469543, + 0.9996563196182251, + 0.9995564222335815, + 0.9996431469917297, + 0.9995015859603882, + 0.9996466636657715, + 0.9994857311248779, + 0.9995725154876709, + 0.9993917942047119, + 0.9997338056564331, + 0.9998142719268799, + 0.9997738003730774, + 0.9997599124908447, + 0.9997965097427368, + 0.9997284412384033, + 0.9997711181640625, + 0.9996548891067505, + 0.999747097492218, + 0.9995355010032654, + 0.9995253086090088, + 0.9992527961730957, + 0.9991543292999268, + 0.9988086819648743, + 0.9985666871070862, + 0.9980965256690979, + 0.9990485310554504, + 0.9984483122825623, + 0.9999804496765137 + ], + "feedback_forward_norm_ratio": [ + 0.9986742734909058, + 1.0006389617919922, + 0.9969250559806824, + 1.0015281438827515, + 0.9994803667068481, + 1.0023949146270752, + 0.9992635250091553, + 1.0001678466796875, + 0.9981926679611206, + 1.0019519329071045, + 0.9987691044807434, + 1.0017606019973755, + 0.9995517730712891, + 1.0003198385238647, + 0.9993669390678406, + 1.0014435052871704, + 0.9996762871742249, + 1.0010688304901123, + 0.9983429312705994, + 1.000935435295105, + 0.9987174272537231, + 1.001394271850586, + 0.9991419315338135, + 1.0014270544052124, + 0.999220609664917, + 1.001044511795044, + 0.9996599555015564, + 1.0019431114196777, + 0.999127209186554, + 1.0018210411071777, + 0.9987342357635498, + 1.002028465270996, + 0.999176025390625, + 1.0014485120773315, + 0.9987326860427856, + 1.0022226572036743, + 0.999302089214325, + 1.0003089904785156, + 0.9998563528060913, + 1.0008293390274048, + 0.9998266696929932, + 1.000866413116455, + 0.9995371103286743, + 1.0013400316238403, + 0.9996753334999084, + 1.0014046430587769, + 0.9987162947654724, + 1.0021916627883911, + 0.9979419708251953, + 1.002637267112732, + 0.9971224069595337, + 1.0037763118743896, + 0.9985743761062622, + 1.0024051666259766, + 1.0020339488983154 + ], + "feedback_forward_relative_error": [ + 0.030373815447092056, + 0.031964294612407684, + 0.03562188893556595, + 0.03850705549120903, + 0.04067551717162132, + 0.04135864973068237, + 0.03340218961238861, + 0.03303094580769539, + 0.03472533077001572, + 0.03634229302406311, + 0.028073109686374664, + 0.030188247561454773, + 0.026667024940252304, + 0.02766183763742447, + 0.02230963297188282, + 0.02419978752732277, + 0.020063288509845734, + 0.025182627141475677, + 0.025750935077667236, + 0.021695468574762344, + 0.02896524965763092, + 0.029999585822224617, + 0.02637401595711708, + 0.029252730309963226, + 0.025553440675139427, + 0.02942599356174469, + 0.024919671937823296, + 0.02895422838628292, + 0.026219619438052177, + 0.029866602271795273, + 0.02672802284359932, + 0.03166596591472626, + 0.02658591978251934, + 0.03212727606296539, + 0.029248615726828575, + 0.03498656675219536, + 0.023075399920344353, + 0.019277160987257957, + 0.0212686900049448, + 0.021937349811196327, + 0.020174607634544373, + 0.023328101262450218, + 0.021395066753029823, + 0.026319433003664017, + 0.02248847484588623, + 0.030529577285051346, + 0.03081503137946129, + 0.038761161267757416, + 0.04113198071718216, + 0.0489472858607769, + 0.053541116416454315, + 0.06193094700574875, + 0.04361523687839508, + 0.05582822114229202, + 0.006581505760550499 + ], + "innovation_negative_gradient_cosine": [ + 0.999534547328949, + 0.999441385269165, + 0.9995296001434326, + 0.9990184307098389, + 0.9995428323745728, + 0.9992603063583374, + 0.9995453953742981, + 0.9992389678955078, + 0.9995508193969727, + 0.9993209838867188, + 0.99955815076828, + 0.9994360208511353, + 0.9995626211166382, + 0.999459981918335, + 0.9995666742324829, + 0.9995467662811279, + 0.9995676875114441, + 0.9995406270027161, + 0.9995849132537842, + 0.9995675086975098, + 0.9996196031570435, + 0.999435544013977, + 0.9996318817138672, + 0.9994937181472778, + 0.9996384382247925, + 0.9995115995407104, + 0.9996570348739624, + 0.9994932413101196, + 0.9996674060821533, + 0.9994916915893555, + 0.9996871948242188, + 0.9995210766792297, + 0.9997069239616394, + 0.9995291829109192, + 0.9997122883796692, + 0.9995144605636597, + 0.9997189044952393, + 0.9997336864471436, + 0.9997418522834778, + 0.9996910095214844, + 0.9997549057006836, + 0.999707043170929, + 0.9997638463973999, + 0.9997535347938538, + 0.9997890591621399, + 0.9998000860214233, + 0.9997377395629883, + 0.9997023344039917, + 0.9997260570526123, + 0.999669075012207, + 0.9996702671051025, + 0.9995840787887573, + 0.9998461008071899, + 0.9998620748519897, + 0.6710851192474365 + ], + "mean_feedback_forward_cosine": 0.9994896823709661, + "mean_feedback_forward_relative_error": 0.03053843260488727, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.999534547328949, + 0.999441385269165, + 0.9995296001434326, + 0.9990184307098389, + 0.9995428323745728, + 0.9992603063583374, + 0.9995453953742981, + 0.9992389678955078, + 0.9995508193969727, + 0.9993209838867188, + 0.99955815076828, + 0.9994360208511353, + 0.9995626211166382, + 0.999459981918335, + 0.9995666742324829, + 0.9995467662811279, + 0.9995676875114441, + 0.9995406270027161, + 0.9995849132537842, + 0.9995675086975098, + 0.9996196031570435, + 0.999435544013977, + 0.9996318817138672, + 0.9994937181472778, + 0.9996384382247925, + 0.9995115995407104, + 0.9996570348739624, + 0.9994932413101196, + 0.9996674060821533, + 0.9994916915893555, + 0.9996871948242188, + 0.9995210766792297, + 0.9997069239616394, + 0.9995291829109192, + 0.9997122883796692, + 0.9995144605636597, + 0.9997189044952393, + 0.9997336864471436, + 0.9997418522834778, + 0.9996910095214844, + 0.9997549057006836, + 0.999707043170929, + 0.9997638463973999, + 0.9997535347938538, + 0.9997890591621399, + 0.9998000860214233, + 0.9997377395629883, + 0.9997023344039917, + 0.9997260570526123, + 0.999669075012207, + 0.9996702671051025, + 0.9995840787887573, + 0.9998461008071899, + 0.9998620748519897, + 0.6710851192474365 + ], + "teaching_negative_gradient_cosine": [ + 0.999534547328949, + 0.999441385269165, + 0.9995296001434326, + 0.9990184307098389, + 0.9995428323745728, + 0.9992603063583374, + 0.9995453953742981, + 0.9992389678955078, + 0.9995508193969727, + 0.9993209838867188, + 0.99955815076828, + 0.9994360208511353, + 0.9995626211166382, + 0.999459981918335, + 0.9995666742324829, + 0.9995467662811279, + 0.9995676875114441, + 0.9995406270027161, + 0.9995849132537842, + 0.9995675086975098, + 0.9996196031570435, + 0.999435544013977, + 0.9996318817138672, + 0.9994937181472778, + 0.9996384382247925, + 0.9995115995407104, + 0.9996570348739624, + 0.9994932413101196, + 0.9996674060821533, + 0.9994916915893555, + 0.9996871948242188, + 0.9995210766792297, + 0.9997069239616394, + 0.9995291829109192, + 0.9997122883796692, + 0.9995144605636597, + 0.9997189044952393, + 0.9997336864471436, + 0.9997418522834778, + 0.9996910095214844, + 0.9997549057006836, + 0.999707043170929, + 0.9997638463973999, + 0.9997535347938538, + 0.9997890591621399, + 0.9998000860214233, + 0.9997377395629883, + 0.9997023344039917, + 0.9997260570526123, + 0.999669075012207, + 0.9996702671051025, + 0.9995840787887573, + 0.9998461008071899, + 0.9998620748519897, + 0.6710851192474365 + ], + "wall_s": 0.1840651035308838 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3464447259902954, + "mean_feedback_forward_cosine": 0.29027721380645577, + "mean_feedback_forward_relative_error": 1.178256191448732, + "min_feedback_forward_cosine": 0.09826618432998657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.3704615573501586 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.342604398727417, + "mean_feedback_forward_cosine": 0.31901492232626133, + "mean_feedback_forward_relative_error": 1.1545552064072002, + "min_feedback_forward_cosine": 0.10470203310251236 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.6967413638687134 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3334531784057617, + "mean_feedback_forward_cosine": 0.3631359176202254, + "mean_feedback_forward_relative_error": 1.1180878097360785, + "min_feedback_forward_cosine": 0.11801935732364655 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4307723350524901 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3191616535186768, + "mean_feedback_forward_cosine": 0.41192043640396814, + "mean_feedback_forward_relative_error": 1.0758537660945546, + "min_feedback_forward_cosine": 0.13609740138053894 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.1842613387298584 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3043830394744873, + "mean_feedback_forward_cosine": 0.4591585630720312, + "mean_feedback_forward_relative_error": 1.0326683066108011, + "min_feedback_forward_cosine": 0.15394139289855957 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.9762814709091187 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2937240600585938, + "mean_feedback_forward_cosine": 0.4997470823201266, + "mean_feedback_forward_relative_error": 0.9934693618254228, + "min_feedback_forward_cosine": 0.1700705885887146 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8191733999633789 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2820374965667725, + "mean_feedback_forward_cosine": 0.5342417137189345, + "mean_feedback_forward_relative_error": 0.9584944692524997, + "min_feedback_forward_cosine": 0.1848730593919754 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7156849175262451 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.270569920539856, + "mean_feedback_forward_cosine": 0.5640160666270689, + "mean_feedback_forward_relative_error": 0.9268558773127469, + "min_feedback_forward_cosine": 0.19986508786678314 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6479228650474549 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.257145881652832, + "mean_feedback_forward_cosine": 0.5917187715118581, + "mean_feedback_forward_relative_error": 0.8961032244292173, + "min_feedback_forward_cosine": 0.2162085771560669 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5960064146232605 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2436530590057373, + "mean_feedback_forward_cosine": 0.6161474466323853, + "mean_feedback_forward_relative_error": 0.867996008829637, + "min_feedback_forward_cosine": 0.23284927010536194 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5511851186561585 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.228519082069397, + "mean_feedback_forward_cosine": 0.6386087173765356, + "mean_feedback_forward_relative_error": 0.8411842422051863, + "min_feedback_forward_cosine": 0.25119492411613464 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5161278319740296 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2122453451156616, + "mean_feedback_forward_cosine": 0.659218141707507, + "mean_feedback_forward_relative_error": 0.8157952037724582, + "min_feedback_forward_cosine": 0.2712697386741638 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.48363601528167727 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1960726976394653, + "mean_feedback_forward_cosine": 0.6787421963431618, + "mean_feedback_forward_relative_error": 0.7907431545582685, + "min_feedback_forward_cosine": 0.2901216745376587 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4618261950588226 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1796756982803345, + "mean_feedback_forward_cosine": 0.6973863617940382, + "mean_feedback_forward_relative_error": 0.7662305073304609, + "min_feedback_forward_cosine": 0.31009507179260254 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4427673635292053 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.165794014930725, + "mean_feedback_forward_cosine": 0.7144045293331146, + "mean_feedback_forward_relative_error": 0.7431726366281509, + "min_feedback_forward_cosine": 0.3328298032283783 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.41954453619003296 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1517013311386108, + "mean_feedback_forward_cosine": 0.7306086350571025, + "mean_feedback_forward_relative_error": 0.7205345245924863, + "min_feedback_forward_cosine": 0.3534374535083771 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4039508693885803 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1365071535110474, + "mean_feedback_forward_cosine": 0.7460741254416379, + "mean_feedback_forward_relative_error": 0.6985243339430202, + "min_feedback_forward_cosine": 0.3746267855167389 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3821706154251099 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1201761960983276, + "mean_feedback_forward_cosine": 0.7606539937582883, + "mean_feedback_forward_relative_error": 0.6771352933211761, + "min_feedback_forward_cosine": 0.3938027620315552 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3685064894771576 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1034852266311646, + "mean_feedback_forward_cosine": 0.7745427044955167, + "mean_feedback_forward_relative_error": 0.6560068325562911, + "min_feedback_forward_cosine": 0.4126436710357666 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3599610644435883 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0848698616027832, + "mean_feedback_forward_cosine": 0.7881048738956451, + "mean_feedback_forward_relative_error": 0.6349188978021795, + "min_feedback_forward_cosine": 0.433307945728302 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3474660382270813 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0681661367416382, + "mean_feedback_forward_cosine": 0.800181284275922, + "mean_feedback_forward_relative_error": 0.6155560723759911, + "min_feedback_forward_cosine": 0.45181435346603394 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3313999942684174 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.04939603805542, + "mean_feedback_forward_cosine": 0.8119089007377625, + "mean_feedback_forward_relative_error": 0.5961991448294033, + "min_feedback_forward_cosine": 0.47210100293159485 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32168210739135744 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0293253660202026, + "mean_feedback_forward_cosine": 0.8231192323294553, + "mean_feedback_forward_relative_error": 0.577244947715239, + "min_feedback_forward_cosine": 0.4929039776325226 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.31462405000686644 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0082814693450928, + "mean_feedback_forward_cosine": 0.8338727604259144, + "mean_feedback_forward_relative_error": 0.558591572263024, + "min_feedback_forward_cosine": 0.5139169692993164 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3024618870449066 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.987269401550293, + "mean_feedback_forward_cosine": 0.8439520521597429, + "mean_feedback_forward_relative_error": 0.54047522761605, + "min_feedback_forward_cosine": 0.534113883972168 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.29681352372169495 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9666692018508911, + "mean_feedback_forward_cosine": 0.8535854989832098, + "mean_feedback_forward_relative_error": 0.5227361795577136, + "min_feedback_forward_cosine": 0.5536783933639526 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.291030342874527 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.946130096912384, + "mean_feedback_forward_cosine": 0.8625394571911205, + "mean_feedback_forward_relative_error": 0.5056248185309496, + "min_feedback_forward_cosine": 0.5723249912261963 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2811413796710968 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9224547743797302, + "mean_feedback_forward_cosine": 0.8712533885782415, + "mean_feedback_forward_relative_error": 0.48866380696946926, + "min_feedback_forward_cosine": 0.5937190055847168 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.27247773029327393 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8997247815132141, + "mean_feedback_forward_cosine": 0.87938373522325, + "mean_feedback_forward_relative_error": 0.4724398217417977, + "min_feedback_forward_cosine": 0.6140031218528748 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.26573392446517946 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8767585158348083, + "mean_feedback_forward_cosine": 0.8873388496312228, + "mean_feedback_forward_relative_error": 0.45595376071604815, + "min_feedback_forward_cosine": 0.6334506869316101 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2651367131471634 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8535721898078918, + "mean_feedback_forward_cosine": 0.8947497736323964, + "mean_feedback_forward_relative_error": 0.4401968778534369, + "min_feedback_forward_cosine": 0.652430534362793 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.261038997964859 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8301363587379456, + "mean_feedback_forward_cosine": 0.9016447034749118, + "mean_feedback_forward_relative_error": 0.4251384423537688, + "min_feedback_forward_cosine": 0.6714699268341064 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.24747045076847077 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8073676228523254, + "mean_feedback_forward_cosine": 0.9081655393947254, + "mean_feedback_forward_relative_error": 0.41030791076746853, + "min_feedback_forward_cosine": 0.6888424158096313 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.24297086081027985 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7856544256210327, + "mean_feedback_forward_cosine": 0.9142789288000627, + "mean_feedback_forward_relative_error": 0.39597937315702436, + "min_feedback_forward_cosine": 0.7051998972892761 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.24031951283931732 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7637181878089905, + "mean_feedback_forward_cosine": 0.9201499971476468, + "mean_feedback_forward_relative_error": 0.3818268209695816, + "min_feedback_forward_cosine": 0.7213693857192993 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.23635845373630524 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7411061525344849, + "mean_feedback_forward_cosine": 0.9256504665721547, + "mean_feedback_forward_relative_error": 0.36806188930164685, + "min_feedback_forward_cosine": 0.7375943660736084 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.23369537858963013 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7196806073188782, + "mean_feedback_forward_cosine": 0.9307226386937228, + "mean_feedback_forward_relative_error": 0.3549185817891901, + "min_feedback_forward_cosine": 0.7523044943809509 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2313125407028198 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6974388957023621, + "mean_feedback_forward_cosine": 0.9354210929437117, + "mean_feedback_forward_relative_error": 0.342401736703786, + "min_feedback_forward_cosine": 0.7674288749694824 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.22255386281967163 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6743432283401489, + "mean_feedback_forward_cosine": 0.939997334913774, + "mean_feedback_forward_relative_error": 0.32986792014403776, + "min_feedback_forward_cosine": 0.7824465036392212 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.223271989402771 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6519086956977844, + "mean_feedback_forward_cosine": 0.9441849589347839, + "mean_feedback_forward_relative_error": 0.3179208685051311, + "min_feedback_forward_cosine": 0.7965607643127441 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.21751073798656465 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6300700306892395, + "mean_feedback_forward_cosine": 0.9482373237609864, + "mean_feedback_forward_relative_error": 0.3059993200681426, + "min_feedback_forward_cosine": 0.809735119342804 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.21728030968666076 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6099371910095215, + "mean_feedback_forward_cosine": 0.9519010370427912, + "mean_feedback_forward_relative_error": 0.29486370913007043, + "min_feedback_forward_cosine": 0.8216880559921265 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.20883438779830932 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5901137590408325, + "mean_feedback_forward_cosine": 0.9552752581509677, + "mean_feedback_forward_relative_error": 0.2841600298881531, + "min_feedback_forward_cosine": 0.8328014612197876 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.20926959754467012 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5698909759521484, + "mean_feedback_forward_cosine": 0.9585013281215321, + "mean_feedback_forward_relative_error": 0.273653789406473, + "min_feedback_forward_cosine": 0.844026505947113 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.20790495951652527 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5506214499473572, + "mean_feedback_forward_cosine": 0.9614172577857971, + "mean_feedback_forward_relative_error": 0.26380565762519836, + "min_feedback_forward_cosine": 0.8544111847877502 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2017436584854126 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5307701230049133, + "mean_feedback_forward_cosine": 0.9642623077739368, + "mean_feedback_forward_relative_error": 0.25384207489815624, + "min_feedback_forward_cosine": 0.8647032976150513 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.19571260312080382 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5120128393173218, + "mean_feedback_forward_cosine": 0.9669614347544584, + "mean_feedback_forward_relative_error": 0.2439642841165716, + "min_feedback_forward_cosine": 0.8739529848098755 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.1975364964056015 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4943452775478363, + "mean_feedback_forward_cosine": 0.9693642713806846, + "mean_feedback_forward_relative_error": 0.23483136296272278, + "min_feedback_forward_cosine": 0.8824831247329712 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.19639045250177384 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4769802987575531, + "mean_feedback_forward_cosine": 0.9715972369367426, + "mean_feedback_forward_relative_error": 0.22603400234471668, + "min_feedback_forward_cosine": 0.8905482292175293 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.18923484493255616 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46077466011047363, + "mean_feedback_forward_cosine": 0.9736243302171881, + "mean_feedback_forward_relative_error": 0.21773624020543966, + "min_feedback_forward_cosine": 0.8977358341217041 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.18611038329124452 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.44443535804748535, + "mean_feedback_forward_cosine": 0.9755336100404913, + "mean_feedback_forward_relative_error": 0.2096899394284595, + "min_feedback_forward_cosine": 0.904762864112854 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.18421099987983702 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42707616090774536, + "mean_feedback_forward_cosine": 0.9773360046473416, + "mean_feedback_forward_relative_error": 0.20185066983103753, + "min_feedback_forward_cosine": 0.9120082855224609 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.18032447082519532 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41098570823669434, + "mean_feedback_forward_cosine": 0.9790595271370628, + "mean_feedback_forward_relative_error": 0.1940204645422372, + "min_feedback_forward_cosine": 0.9184935092926025 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.1868881555366516 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3946766257286072, + "mean_feedback_forward_cosine": 0.9806340380148454, + "mean_feedback_forward_relative_error": 0.18661480302160435, + "min_feedback_forward_cosine": 0.9246987104415894 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.17931044955730438 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37967368960380554, + "mean_feedback_forward_cosine": 0.9820866823196411, + "mean_feedback_forward_relative_error": 0.17948700020259076, + "min_feedback_forward_cosine": 0.9302462339401245 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1762636404657364 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36562249064445496, + "mean_feedback_forward_cosine": 0.9834293506362222, + "mean_feedback_forward_relative_error": 0.17262107696045528, + "min_feedback_forward_cosine": 0.9352699518203735 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1766922233390808 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35152167081832886, + "mean_feedback_forward_cosine": 0.9846597454764626, + "mean_feedback_forward_relative_error": 0.16610525501045315, + "min_feedback_forward_cosine": 0.9401300549507141 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.17598136737823486 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3375950753688812, + "mean_feedback_forward_cosine": 0.9858143698085439, + "mean_feedback_forward_relative_error": 0.15978123186664148, + "min_feedback_forward_cosine": 0.9447086453437805 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.17427285524368286 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3254486322402954, + "mean_feedback_forward_cosine": 0.9868530143391002, + "mean_feedback_forward_relative_error": 0.15380413701588458, + "min_feedback_forward_cosine": 0.94859379529953 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.16943655662059784 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3130016028881073, + "mean_feedback_forward_cosine": 0.9878478624603965, + "mean_feedback_forward_relative_error": 0.14787317527966065, + "min_feedback_forward_cosine": 0.9523912668228149 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.17375968336105346 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3006737530231476, + "mean_feedback_forward_cosine": 0.9887408408251676, + "mean_feedback_forward_relative_error": 0.14235474067655476, + "min_feedback_forward_cosine": 0.9560116529464722 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1664853154194355 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2892468273639679, + "mean_feedback_forward_cosine": 0.9895924654873934, + "mean_feedback_forward_relative_error": 0.13685577057979323, + "min_feedback_forward_cosine": 0.9592477083206177 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.16777808388233184 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2781123220920563, + "mean_feedback_forward_cosine": 0.9903776082125577, + "mean_feedback_forward_relative_error": 0.13159775713627989, + "min_feedback_forward_cosine": 0.9622933864593506 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.16732357576370238 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2673693597316742, + "mean_feedback_forward_cosine": 0.9910991148515181, + "mean_feedback_forward_relative_error": 0.12656967727975413, + "min_feedback_forward_cosine": 0.9651110172271729 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1668574471092224 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2565230429172516, + "mean_feedback_forward_cosine": 0.9917782989415256, + "mean_feedback_forward_relative_error": 0.12167377007955854, + "min_feedback_forward_cosine": 0.9678463339805603 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1682041015291214 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.246418297290802, + "mean_feedback_forward_cosine": 0.9923990683122115, + "mean_feedback_forward_relative_error": 0.11701159981841391, + "min_feedback_forward_cosine": 0.9703002572059631 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.16243797614574432 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23698997497558594, + "mean_feedback_forward_cosine": 0.9929798906499689, + "mean_feedback_forward_relative_error": 0.11245247261090711, + "min_feedback_forward_cosine": 0.9725117683410645 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.16206168502807616 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22819016873836517, + "mean_feedback_forward_cosine": 0.9934903372417797, + "mean_feedback_forward_relative_error": 0.10829746635122732, + "min_feedback_forward_cosine": 0.9744926691055298 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1601631025314331 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2193627506494522, + "mean_feedback_forward_cosine": 0.9939808195287531, + "mean_feedback_forward_relative_error": 0.10415428670292551, + "min_feedback_forward_cosine": 0.9764065742492676 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.15879637899398805 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21048295497894287, + "mean_feedback_forward_cosine": 0.9944382244890386, + "mean_feedback_forward_relative_error": 0.10016306428746743, + "min_feedback_forward_cosine": 0.9782607555389404 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.160089029340744 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20216532051563263, + "mean_feedback_forward_cosine": 0.9948591644113715, + "mean_feedback_forward_relative_error": 0.0963318647308783, + "min_feedback_forward_cosine": 0.9799298644065857 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.15733842891693114 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19413991272449493, + "mean_feedback_forward_cosine": 0.995243741165508, + "mean_feedback_forward_relative_error": 0.09266701493073594, + "min_feedback_forward_cosine": 0.9814789891242981 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.15482519105911255 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18668678402900696, + "mean_feedback_forward_cosine": 0.9955994064157659, + "mean_feedback_forward_relative_error": 0.08915054053068161, + "min_feedback_forward_cosine": 0.9828635454177856 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1533050349807739 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17943738400936127, + "mean_feedback_forward_cosine": 0.9959320068359375, + "mean_feedback_forward_relative_error": 0.08572588434273547, + "min_feedback_forward_cosine": 0.9841609001159668 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.15046897461891173 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17239177227020264, + "mean_feedback_forward_cosine": 0.99624718319286, + "mean_feedback_forward_relative_error": 0.08235500065440482, + "min_feedback_forward_cosine": 0.9853790402412415 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.15597980010747908 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16524744033813477, + "mean_feedback_forward_cosine": 0.9965404543009672, + "mean_feedback_forward_relative_error": 0.07910626672885634, + "min_feedback_forward_cosine": 0.9865565299987793 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.15417349669456482 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15881438553333282, + "mean_feedback_forward_cosine": 0.9967974153431979, + "mean_feedback_forward_relative_error": 0.07613672224635427, + "min_feedback_forward_cosine": 0.9875755906105042 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1473866323184967 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15201766788959503, + "mean_feedback_forward_cosine": 0.9970540989529003, + "mean_feedback_forward_relative_error": 0.0730514292689887, + "min_feedback_forward_cosine": 0.988610029220581 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15314912130355834 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14590775966644287, + "mean_feedback_forward_cosine": 0.9972771482034163, + "mean_feedback_forward_relative_error": 0.07023630430075255, + "min_feedback_forward_cosine": 0.9895028471946716 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15089951363563536 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13983440399169922, + "mean_feedback_forward_cosine": 0.9974843111905185, + "mean_feedback_forward_relative_error": 0.06753264333714139, + "min_feedback_forward_cosine": 0.9903545379638672 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15078624309539795 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13432933390140533, + "mean_feedback_forward_cosine": 0.9976726553656838, + "mean_feedback_forward_relative_error": 0.06496261497112837, + "min_feedback_forward_cosine": 0.9910942316055298 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.14839280843496322 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1289684921503067, + "mean_feedback_forward_cosine": 0.9978530916300687, + "mean_feedback_forward_relative_error": 0.062406730838119984, + "min_feedback_forward_cosine": 0.9917852878570557 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.14823583701610565 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12417666614055634, + "mean_feedback_forward_cosine": 0.9980134508826516, + "mean_feedback_forward_relative_error": 0.06003500163893808, + "min_feedback_forward_cosine": 0.9923818111419678 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.14464116917133332 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11920725554227829, + "mean_feedback_forward_cosine": 0.9981636946851556, + "mean_feedback_forward_relative_error": 0.05773626547306776, + "min_feedback_forward_cosine": 0.9929776191711426 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.14492770351409912 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11448901146650314, + "mean_feedback_forward_cosine": 0.99830322265625, + "mean_feedback_forward_relative_error": 0.05551600430838086, + "min_feedback_forward_cosine": 0.9935189485549927 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.14268309606552124 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10983197391033173, + "mean_feedback_forward_cosine": 0.9984360218048096, + "mean_feedback_forward_relative_error": 0.053309309702705254, + "min_feedback_forward_cosine": 0.9940335750579834 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1478362784576416 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10562899708747864, + "mean_feedback_forward_cosine": 0.9985541972247037, + "mean_feedback_forward_relative_error": 0.05125798300247301, + "min_feedback_forward_cosine": 0.9944791793823242 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1437551615524292 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10163402557373047, + "mean_feedback_forward_cosine": 0.9986639445478266, + "mean_feedback_forward_relative_error": 0.04928394603458318, + "min_feedback_forward_cosine": 0.9948862791061401 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14351468017578126 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09729874134063721, + "mean_feedback_forward_cosine": 0.9987684997645292, + "mean_feedback_forward_relative_error": 0.047331926243549045, + "min_feedback_forward_cosine": 0.9953108429908752 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.14870515768051148 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09344491362571716, + "mean_feedback_forward_cosine": 0.9988601771267978, + "mean_feedback_forward_relative_error": 0.04554461816495115, + "min_feedback_forward_cosine": 0.9956722259521484 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.14165220482349397 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08984030038118362, + "mean_feedback_forward_cosine": 0.9989478772336786, + "mean_feedback_forward_relative_error": 0.043766251460395075, + "min_feedback_forward_cosine": 0.995998740196228 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1420630505466461 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08653393387794495, + "mean_feedback_forward_cosine": 0.9990258845415982, + "mean_feedback_forward_relative_error": 0.04212138226086443, + "min_feedback_forward_cosine": 0.9962879419326782 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.13872229014873505 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08304337412118912, + "mean_feedback_forward_cosine": 0.9991016897288236, + "mean_feedback_forward_relative_error": 0.04045714531093836, + "min_feedback_forward_cosine": 0.9965806603431702 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1445293878555298 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07964146882295609, + "mean_feedback_forward_cosine": 0.9991688966751099, + "mean_feedback_forward_relative_error": 0.03892969995398413, + "min_feedback_forward_cosine": 0.9968552589416504 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.13843780771970748 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07634511590003967, + "mean_feedback_forward_cosine": 0.9992327451705932, + "mean_feedback_forward_relative_error": 0.03741777705197984, + "min_feedback_forward_cosine": 0.9971089363098145 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.13654371682167052 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07344254851341248, + "mean_feedback_forward_cosine": 0.9992919184944846, + "mean_feedback_forward_relative_error": 0.03594837166707624, + "min_feedback_forward_cosine": 0.9973240494728088 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1398654334306717 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07045339047908783, + "mean_feedback_forward_cosine": 0.9993484518744729, + "mean_feedback_forward_relative_error": 0.03449128752743656, + "min_feedback_forward_cosine": 0.9975365996360779 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14294681231498718 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06773552298545837, + "mean_feedback_forward_cosine": 0.9993969202041626, + "mean_feedback_forward_relative_error": 0.033191636018455026, + "min_feedback_forward_cosine": 0.9977228045463562 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.13541974694252015 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06507350504398346, + "mean_feedback_forward_cosine": 0.9994427128271623, + "mean_feedback_forward_relative_error": 0.031913321566852657, + "min_feedback_forward_cosine": 0.9978986978530884 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14026688925266265 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0627213791012764, + "mean_feedback_forward_cosine": 0.9994821851903742, + "mean_feedback_forward_relative_error": 0.030764484193853355, + "min_feedback_forward_cosine": 0.9980470538139343 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.13205082747459412 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06269843876361847, + "mean_feedback_forward_cosine": 0.9994824994694103, + "mean_feedback_forward_relative_error": 0.03075404701415788, + "min_feedback_forward_cosine": 0.9980485439300537 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.06759102570533752 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06268610805273056, + "mean_feedback_forward_cosine": 0.9994826923717152, + "mean_feedback_forward_relative_error": 0.03074822404675863, + "min_feedback_forward_cosine": 0.9980494976043701 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.045256944432854655 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06267379224300385, + "mean_feedback_forward_cosine": 0.9994828408414668, + "mean_feedback_forward_relative_error": 0.03074289435012774, + "min_feedback_forward_cosine": 0.9980503916740417 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04034666366100311 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06266102194786072, + "mean_feedback_forward_cosine": 0.9994830001484264, + "mean_feedback_forward_relative_error": 0.030737932263450188, + "min_feedback_forward_cosine": 0.9980511665344238 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03293282743215561 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06264709681272507, + "mean_feedback_forward_cosine": 0.9994831702925943, + "mean_feedback_forward_relative_error": 0.030732802640308032, + "min_feedback_forward_cosine": 0.9980521202087402 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03022145635008812 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06263425946235657, + "mean_feedback_forward_cosine": 0.9994833133437416, + "mean_feedback_forward_relative_error": 0.030728215766562657, + "min_feedback_forward_cosine": 0.9980529546737671 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.025835701021552086 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06262177228927612, + "mean_feedback_forward_cosine": 0.999483428218148, + "mean_feedback_forward_relative_error": 0.030723788335242055, + "min_feedback_forward_cosine": 0.9980536699295044 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.02356987555027008 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06260990351438522, + "mean_feedback_forward_cosine": 0.9994835582646456, + "mean_feedback_forward_relative_error": 0.03071964384818619, + "min_feedback_forward_cosine": 0.998054563999176 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.021534102993607522 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06259550154209137, + "mean_feedback_forward_cosine": 0.9994837110692805, + "mean_feedback_forward_relative_error": 0.03071519649502906, + "min_feedback_forward_cosine": 0.9980553388595581 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.020321601420640944 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0625816360116005, + "mean_feedback_forward_cosine": 0.9994838573715904, + "mean_feedback_forward_relative_error": 0.03071072819558057, + "min_feedback_forward_cosine": 0.9980563521385193 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.019604212568700312 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06256827712059021, + "mean_feedback_forward_cosine": 0.9994839744134383, + "mean_feedback_forward_relative_error": 0.03070665564049374, + "min_feedback_forward_cosine": 0.9980570673942566 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.016796411416381597 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06255496293306351, + "mean_feedback_forward_cosine": 0.9994841022924943, + "mean_feedback_forward_relative_error": 0.03070269122042439, + "min_feedback_forward_cosine": 0.9980579614639282 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.015727011843919753 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06253934651613235, + "mean_feedback_forward_cosine": 0.9994842648506165, + "mean_feedback_forward_relative_error": 0.030698163578794763, + "min_feedback_forward_cosine": 0.998059093952179 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.015990155968666077 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06252351403236389, + "mean_feedback_forward_cosine": 0.9994844100692055, + "mean_feedback_forward_relative_error": 0.03069376213984056, + "min_feedback_forward_cosine": 0.9980600476264954 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01549170794636011 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06250815838575363, + "mean_feedback_forward_cosine": 0.9994845433668657, + "mean_feedback_forward_relative_error": 0.030689492339098994, + "min_feedback_forward_cosine": 0.9980608820915222 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.014437230432629585 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06249437481164932, + "mean_feedback_forward_cosine": 0.9994846593249928, + "mean_feedback_forward_relative_error": 0.030685534599152478, + "min_feedback_forward_cosine": 0.9980616569519043 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.01300610273450613 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062480244785547256, + "mean_feedback_forward_cosine": 0.9994848132133484, + "mean_feedback_forward_relative_error": 0.03068151062523777, + "min_feedback_forward_cosine": 0.9980626106262207 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01203647702410817 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0624653697013855, + "mean_feedback_forward_cosine": 0.9994849421761253, + "mean_feedback_forward_relative_error": 0.030677358492870222, + "min_feedback_forward_cosine": 0.9980636239051819 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.012282160293906927 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06244941055774689, + "mean_feedback_forward_cosine": 0.9994850819761103, + "mean_feedback_forward_relative_error": 0.030673144935545595, + "min_feedback_forward_cosine": 0.9980645775794983 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.011743366521783174 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062433239072561264, + "mean_feedback_forward_cosine": 0.9994852185249329, + "mean_feedback_forward_relative_error": 0.03066888998516581, + "min_feedback_forward_cosine": 0.9980655312538147 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.011445523285418749 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06241831183433533, + "mean_feedback_forward_cosine": 0.9994853594086387, + "mean_feedback_forward_relative_error": 0.03066485823047432, + "min_feedback_forward_cosine": 0.9980664849281311 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.010189418796002865 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06240494176745415, + "mean_feedback_forward_cosine": 0.9994854992086237, + "mean_feedback_forward_relative_error": 0.030661224785514852, + "min_feedback_forward_cosine": 0.9980674386024475 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.009600654789209366 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0623897947371006, + "mean_feedback_forward_cosine": 0.9994856108318676, + "mean_feedback_forward_relative_error": 0.030657286281612788, + "min_feedback_forward_cosine": 0.9980683326721191 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.009567934661507606 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062374573200941086, + "mean_feedback_forward_cosine": 0.9994857527992942, + "mean_feedback_forward_relative_error": 0.03065339655361392, + "min_feedback_forward_cosine": 0.998069167137146 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.009035898718833923 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06235916540026665, + "mean_feedback_forward_cosine": 0.9994858785109086, + "mean_feedback_forward_relative_error": 0.030649371428245847, + "min_feedback_forward_cosine": 0.998070240020752 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.00903761274561286 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06234424188733101, + "mean_feedback_forward_cosine": 0.9994860139760104, + "mean_feedback_forward_relative_error": 0.030645305807278915, + "min_feedback_forward_cosine": 0.9980710744857788 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.008683195413649081 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06232771649956703, + "mean_feedback_forward_cosine": 0.9994861537759955, + "mean_feedback_forward_relative_error": 0.030641083588654344, + "min_feedback_forward_cosine": 0.9980721473693848 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.008440109309703112 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062310799956321716, + "mean_feedback_forward_cosine": 0.9994863054969094, + "mean_feedback_forward_relative_error": 0.030636796067384155, + "min_feedback_forward_cosine": 0.998073160648346 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.008790008682310581 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06229575350880623, + "mean_feedback_forward_cosine": 0.9994864637201483, + "mean_feedback_forward_relative_error": 0.030632674372331664, + "min_feedback_forward_cosine": 0.998073935508728 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.007810075041279197 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062281206250190735, + "mean_feedback_forward_cosine": 0.9994865883480418, + "mean_feedback_forward_relative_error": 0.03062878699803894, + "min_feedback_forward_cosine": 0.9980750679969788 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.007518504073172808 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06226494535803795, + "mean_feedback_forward_cosine": 0.999486716227098, + "mean_feedback_forward_relative_error": 0.030624614342708478, + "min_feedback_forward_cosine": 0.9980760216712952 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.007846000881716608 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062249086797237396, + "mean_feedback_forward_cosine": 0.999486861445687, + "mean_feedback_forward_relative_error": 0.030620227356187323, + "min_feedback_forward_cosine": 0.9980770349502563 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007394560450278222 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06223296746611595, + "mean_feedback_forward_cosine": 0.9994870283386924, + "mean_feedback_forward_relative_error": 0.030616129477593033, + "min_feedback_forward_cosine": 0.9980779886245728 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.00689208334473893 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062215469777584076, + "mean_feedback_forward_cosine": 0.9994871757247231, + "mean_feedback_forward_relative_error": 0.0306115483018485, + "min_feedback_forward_cosine": 0.9980791807174683 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.00785630393801257 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062198810279369354, + "mean_feedback_forward_cosine": 0.9994873187758706, + "mean_feedback_forward_relative_error": 0.030607500587674703, + "min_feedback_forward_cosine": 0.9980801343917847 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.007258220757246018 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06218484044075012, + "mean_feedback_forward_cosine": 0.9994874531572515, + "mean_feedback_forward_relative_error": 0.030603864069350742, + "min_feedback_forward_cosine": 0.9980810880661011 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.0061892494452372195 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06217186897993088, + "mean_feedback_forward_cosine": 0.9994875712828203, + "mean_feedback_forward_relative_error": 0.030600314316424456, + "min_feedback_forward_cosine": 0.9980817437171936 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.0059076879094541075 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06215699762105942, + "mean_feedback_forward_cosine": 0.9994876991618763, + "mean_feedback_forward_relative_error": 0.03059661764312874, + "min_feedback_forward_cosine": 0.99808269739151 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.005383817988932132 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062138814479112625, + "mean_feedback_forward_cosine": 0.9994878508827904, + "mean_feedback_forward_relative_error": 0.030592034469273957, + "min_feedback_forward_cosine": 0.9980838894844055 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.006710900447592139 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06212152913212776, + "mean_feedback_forward_cosine": 0.9994880058548667, + "mean_feedback_forward_relative_error": 0.03058764510364695, + "min_feedback_forward_cosine": 0.9980849623680115 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.006575945821553469 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06210213527083397, + "mean_feedback_forward_cosine": 0.9994881868362426, + "mean_feedback_forward_relative_error": 0.030582719304683533, + "min_feedback_forward_cosine": 0.998086154460907 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.006847478371560574 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0620831735432148, + "mean_feedback_forward_cosine": 0.9994883591478522, + "mean_feedback_forward_relative_error": 0.030578120323744688, + "min_feedback_forward_cosine": 0.9980872869491577 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.00634564375422895 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06206458806991577, + "mean_feedback_forward_cosine": 0.9994885032827204, + "mean_feedback_forward_relative_error": 0.0305734507993541, + "min_feedback_forward_cosine": 0.9980883598327637 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.006293262675926089 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062043607234954834, + "mean_feedback_forward_cosine": 0.9994886961850253, + "mean_feedback_forward_relative_error": 0.03056816380809654, + "min_feedback_forward_cosine": 0.9980897307395935 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006856801314167678 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06202723830938339, + "mean_feedback_forward_cosine": 0.9994888533245433, + "mean_feedback_forward_relative_error": 0.030563697379759766, + "min_feedback_forward_cosine": 0.9980907440185547 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.005869663535952568 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06201047822833061, + "mean_feedback_forward_cosine": 0.9994889909570868, + "mean_feedback_forward_relative_error": 0.030559722884473475, + "min_feedback_forward_cosine": 0.9980918169021606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.005216972618214786 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06199338287115097, + "mean_feedback_forward_cosine": 0.9994891394268383, + "mean_feedback_forward_relative_error": 0.030555458688600497, + "min_feedback_forward_cosine": 0.9980928301811218 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.005385983447609469 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06197473779320717, + "mean_feedback_forward_cosine": 0.999489304152402, + "mean_feedback_forward_relative_error": 0.03055039835585789, + "min_feedback_forward_cosine": 0.998093843460083 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.005695979114994407 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06195850670337677, + "mean_feedback_forward_cosine": 0.999489455873316, + "mean_feedback_forward_relative_error": 0.030546232837844977, + "min_feedback_forward_cosine": 0.9980949759483337 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.005280598983112722 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061936236917972565, + "mean_feedback_forward_cosine": 0.9994896509430625, + "mean_feedback_forward_relative_error": 0.030540636054832825, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.00612729561381042 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193609535694122, + "mean_feedback_forward_cosine": 0.9994896455244584, + "mean_feedback_forward_relative_error": 0.03054057719524611, + "min_feedback_forward_cosine": 0.998096227645874 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.004530388177447021 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193596124649048, + "mean_feedback_forward_cosine": 0.999489659612829, + "mean_feedback_forward_relative_error": 0.030540518132461742, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004308553957641125 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061935849487781525, + "mean_feedback_forward_cosine": 0.9994896574453874, + "mean_feedback_forward_relative_error": 0.030540465174073522, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.004125927190994844 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061935700476169586, + "mean_feedback_forward_cosine": 0.9994896509430625, + "mean_feedback_forward_relative_error": 0.030540410141375932, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.00451846117965877 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193555146455765, + "mean_feedback_forward_cosine": 0.9994896617802707, + "mean_feedback_forward_relative_error": 0.03054035593840209, + "min_feedback_forward_cosine": 0.9980962872505188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004326087524071336 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061935409903526306, + "mean_feedback_forward_cosine": 0.9994896531105042, + "mean_feedback_forward_relative_error": 0.030540305858647282, + "min_feedback_forward_cosine": 0.998096227645874 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004024175264686346 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193528696894646, + "mean_feedback_forward_cosine": 0.9994896563616666, + "mean_feedback_forward_relative_error": 0.0305402558804913, + "min_feedback_forward_cosine": 0.998096227645874 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.003828567948546261 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193516030907631, + "mean_feedback_forward_cosine": 0.9994896617802707, + "mean_feedback_forward_relative_error": 0.03054020767184821, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.004268716067522764 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193503737449646, + "mean_feedback_forward_cosine": 0.9994896693663163, + "mean_feedback_forward_relative_error": 0.030540157355029473, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0038033735569939016 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193489581346512, + "mean_feedback_forward_cosine": 0.9994896802035245, + "mean_feedback_forward_relative_error": 0.03054010575975884, + "min_feedback_forward_cosine": 0.9980962872505188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.003953896807860583 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193477660417557, + "mean_feedback_forward_cosine": 0.9994896704500372, + "mean_feedback_forward_relative_error": 0.030540055095810783, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0038096044862642882 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061934683471918106, + "mean_feedback_forward_cosine": 0.9994896758686412, + "mean_feedback_forward_relative_error": 0.030540014862675557, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.0033296207774430515 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061934567987918854, + "mean_feedback_forward_cosine": 0.9994896531105042, + "mean_feedback_forward_relative_error": 0.03053997020152482, + "min_feedback_forward_cosine": 0.9980962872505188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0036760004641488197 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0619344562292099, + "mean_feedback_forward_cosine": 0.9994896693663163, + "mean_feedback_forward_relative_error": 0.0305399274114858, + "min_feedback_forward_cosine": 0.998096227645874 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.003656223715506494 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193436309695244, + "mean_feedback_forward_cosine": 0.9994896780360829, + "mean_feedback_forward_relative_error": 0.030539883520792833, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.003014061942361295 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061934247612953186, + "mean_feedback_forward_cosine": 0.9994896661151539, + "mean_feedback_forward_relative_error": 0.03053983548148112, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003544073878750205 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193413957953453, + "mean_feedback_forward_cosine": 0.9994896639477123, + "mean_feedback_forward_relative_error": 0.030539789008484645, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003221760457754135 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193403899669647, + "mean_feedback_forward_cosine": 0.9994896747849205, + "mean_feedback_forward_relative_error": 0.03053974864835089, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.003004233146267943 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193394958972931, + "mean_feedback_forward_cosine": 0.9994896628639914, + "mean_feedback_forward_relative_error": 0.030539709736000408, + "min_feedback_forward_cosine": 0.9980964064598083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0030014431909099223 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061933841556310654, + "mean_feedback_forward_cosine": 0.9994896726174788, + "mean_feedback_forward_relative_error": 0.030539668173613872, + "min_feedback_forward_cosine": 0.9980962872505188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0033193327287212014 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0619337260723114, + "mean_feedback_forward_cosine": 0.9994896682825956, + "mean_feedback_forward_relative_error": 0.030539624198255212, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0032627053089253604 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193362548947334, + "mean_feedback_forward_cosine": 0.9994896726174788, + "mean_feedback_forward_relative_error": 0.030539582204073667, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0028701702140644194 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193353980779648, + "mean_feedback_forward_cosine": 0.9994896563616666, + "mean_feedback_forward_relative_error": 0.03053954222493551, + "min_feedback_forward_cosine": 0.9980962872505188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0028575154671818016 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193346157670021, + "mean_feedback_forward_cosine": 0.9994896780360829, + "mean_feedback_forward_relative_error": 0.030539504176175052, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0030732198222354055 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061933357268571854, + "mean_feedback_forward_cosine": 0.9994896802035245, + "mean_feedback_forward_relative_error": 0.030539463240314615, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0028843802531808614 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061933282762765884, + "mean_feedback_forward_cosine": 0.9994896758686412, + "mean_feedback_forward_relative_error": 0.03053942309184508, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.002822302868962288 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193319708108902, + "mean_feedback_forward_cosine": 0.9994896747849205, + "mean_feedback_forward_relative_error": 0.03053938172418963, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0028555231500044464 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061933111399412155, + "mean_feedback_forward_cosine": 0.9994896856221286, + "mean_feedback_forward_relative_error": 0.030539340161803093, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.002892176869865507 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193302571773529, + "mean_feedback_forward_cosine": 0.9994896758686412, + "mean_feedback_forward_relative_error": 0.030539298768747936, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0029899260829761624 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193292886018753, + "mean_feedback_forward_cosine": 0.9994896780360829, + "mean_feedback_forward_relative_error": 0.030539256740700117, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.002768810143917799 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193281337618828, + "mean_feedback_forward_cosine": 0.9994896802035245, + "mean_feedback_forward_relative_error": 0.03053921466185288, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.003306742629110813 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061932727694511414, + "mean_feedback_forward_cosine": 0.9994896932081743, + "mean_feedback_forward_relative_error": 0.030539175173775718, + "min_feedback_forward_cosine": 0.9980965256690979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0029128653236106038 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061932627111673355, + "mean_feedback_forward_cosine": 0.9994896953756159, + "mean_feedback_forward_relative_error": 0.030539133357392115, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.002987928712326102 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193254888057709, + "mean_feedback_forward_cosine": 0.9994896812872454, + "mean_feedback_forward_relative_error": 0.03053909654475071, + "min_feedback_forward_cosine": 0.9980963468551636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0026746951858513056 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193244829773903, + "mean_feedback_forward_cosine": 0.9994896877895701, + "mean_feedback_forward_relative_error": 0.030539051570336928, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0030044303303956984 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193232908844948, + "mean_feedback_forward_cosine": 0.9994896932081743, + "mean_feedback_forward_relative_error": 0.03053900718858296, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0030119929092749954 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193223595619202, + "mean_feedback_forward_cosine": 0.99948970079422, + "mean_feedback_forward_relative_error": 0.030538960571654818, + "min_feedback_forward_cosine": 0.9980964064598083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.003049773293584585 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193213537335396, + "mean_feedback_forward_cosine": 0.9994896953756159, + "mean_feedback_forward_relative_error": 0.03053891994059086, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0029004252321086824 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0619320422410965, + "mean_feedback_forward_cosine": 0.9994896791198037, + "mean_feedback_forward_relative_error": 0.030538882730020717, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0028327848418429496 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193194538354874, + "mean_feedback_forward_cosine": 0.9994896986267784, + "mean_feedback_forward_relative_error": 0.030538840193978765, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0027145167157426476 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193183735013008, + "mean_feedback_forward_cosine": 0.9994897051291032, + "mean_feedback_forward_relative_error": 0.030538800832900136, + "min_feedback_forward_cosine": 0.9980964064598083 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0029637707200273872 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061931759119033813, + "mean_feedback_forward_cosine": 0.9994896899570118, + "mean_feedback_forward_relative_error": 0.030538763639263132, + "min_feedback_forward_cosine": 0.9980966448783875 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.002693286809436977 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061931658536195755, + "mean_feedback_forward_cosine": 0.9994897029616616, + "mean_feedback_forward_relative_error": 0.030538724193518812, + "min_feedback_forward_cosine": 0.9980966448783875 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.002560252621918917 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0619315542280674, + "mean_feedback_forward_cosine": 0.9994897051291032, + "mean_feedback_forward_relative_error": 0.03053867809305137, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.002967231141254306 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.061931461095809937, + "mean_feedback_forward_cosine": 0.9994897094639865, + "mean_feedback_forward_relative_error": 0.030538642626594413, + "min_feedback_forward_cosine": 0.9980966448783875 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.002552174325212836 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193136051297188, + "mean_feedback_forward_cosine": 0.9994896986267784, + "mean_feedback_forward_relative_error": 0.030538603087717837, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0028820837849471716 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193125247955322, + "mean_feedback_forward_cosine": 0.9994897051291032, + "mean_feedback_forward_relative_error": 0.030538562515919857, + "min_feedback_forward_cosine": 0.9980964660644531 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.002910595020055771 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193114072084427, + "mean_feedback_forward_cosine": 0.9994897094639865, + "mean_feedback_forward_relative_error": 0.03053851816803217, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.00310367449041456 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193104013800621, + "mean_feedback_forward_cosine": 0.9994897072965448, + "mean_feedback_forward_relative_error": 0.03053847612305121, + "min_feedback_forward_cosine": 0.9980965256690979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0028941941729560493 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06193094700574875, + "mean_feedback_forward_cosine": 0.9994896823709661, + "mean_feedback_forward_relative_error": 0.03053843260488727, + "min_feedback_forward_cosine": 0.9980965256690979 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0028094120770692827 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9287, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3952233123779297, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3318545920, + "peak_memory_reserved_bytes": 4754243584, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0534486770629883, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 18251.1642100811, + "train_wall_s": 18245.253133296967 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250433280000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 5010580480000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json new file mode 100644 index 0000000..43d8e81 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json @@ -0,0 +1,3654 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9994188480907016, + "feedback_forward_cosine": [ + 0.9996993541717529, + 0.9996830821037292, + 0.9995632767677307, + 0.9994942545890808, + 0.9996981620788574, + 0.999686598777771, + 0.9995149374008179, + 0.9995197057723999, + 0.9996012449264526, + 0.999535083770752, + 0.9995328187942505, + 0.9994897246360779, + 0.9996421933174133, + 0.9995424151420593, + 0.9996671676635742, + 0.999509334564209, + 0.9997736215591431, + 0.9996488690376282, + 0.9997099041938782, + 0.9997872114181519, + 0.9996638298034668, + 0.9996258020401001, + 0.9996758103370667, + 0.9996204376220703, + 0.9996803402900696, + 0.9995815753936768, + 0.9996604323387146, + 0.9995313882827759, + 0.9996680617332458, + 0.9995308518409729, + 0.9996833801269531, + 0.9995642304420471, + 0.9996212720870972, + 0.9994816780090332, + 0.999650776386261, + 0.9994839429855347, + 0.9997327923774719, + 0.9998137354850769, + 0.9997699856758118, + 0.9997382760047913, + 0.9997545480728149, + 0.9996718764305115, + 0.999742865562439, + 0.9995942115783691, + 0.9997011423110962, + 0.9995160102844238, + 0.9994035959243774, + 0.9991016983985901, + 0.9994409084320068, + 0.999116063117981, + 0.9989480972290039, + 0.9984638690948486, + 0.9990649223327637, + 0.998519241809845, + 0.9999802708625793 + ], + "feedback_forward_norm_ratio": [ + 0.9994032382965088, + 1.0002179145812988, + 0.9992724061012268, + 1.0009822845458984, + 0.9994022846221924, + 1.0009336471557617, + 0.9990427494049072, + 1.0013375282287598, + 0.9988051056861877, + 1.0024958848953247, + 0.9994823336601257, + 1.002917766571045, + 1.0000933408737183, + 1.00282621383667, + 0.9992473721504211, + 1.0023692846298218, + 0.9999445676803589, + 1.0007997751235962, + 0.99896240234375, + 1.0006413459777832, + 0.9993771314620972, + 1.0014231204986572, + 0.9992647171020508, + 1.0011144876480103, + 0.999278724193573, + 1.0012441873550415, + 0.9986801743507385, + 1.0012577772140503, + 0.9993001222610474, + 1.0015628337860107, + 0.9993846416473389, + 1.0021334886550903, + 0.9988804459571838, + 1.0015361309051514, + 0.9989018440246582, + 1.0020201206207275, + 0.9992831349372864, + 1.0004626512527466, + 0.9996106624603271, + 1.0006550550460815, + 0.9993950128555298, + 1.0010838508605957, + 0.9993111491203308, + 1.001451849937439, + 0.9995183348655701, + 1.001678228378296, + 0.9984020590782166, + 1.0025371313095093, + 0.9987742304801941, + 1.0025242567062378, + 0.9982123970985413, + 1.0027648210525513, + 0.9986278414726257, + 1.0024317502975464, + 1.0019079446792603 + ], + "feedback_forward_relative_error": [ + 0.02452433668076992, + 0.025184625759720802, + 0.02955569326877594, + 0.03183719143271446, + 0.02456962876021862, + 0.02506248652935028, + 0.0311441570520401, + 0.031041115522384644, + 0.0282449834048748, + 0.030632462352514267, + 0.03056534379720688, + 0.032126203179359436, + 0.026748819276690483, + 0.030421754345297813, + 0.0258016437292099, + 0.031452398747205734, + 0.021281296387314796, + 0.026522092521190643, + 0.024094196036458015, + 0.020647048950195312, + 0.025927532464265823, + 0.027411917224526405, + 0.0254663098603487, + 0.027587715536355972, + 0.02528620883822441, + 0.028972430154681206, + 0.026078246533870697, + 0.030659740790724754, + 0.02576402947306633, + 0.03069719485938549, + 0.02516496740281582, + 0.02963043749332428, + 0.027530504390597343, + 0.03225620836019516, + 0.026434797793626785, + 0.03222276642918587, + 0.02312055230140686, + 0.019309254363179207, + 0.021448656916618347, + 0.022896774113178253, + 0.022158363834023476, + 0.0256564412266016, + 0.02268141880631447, + 0.028542323037981987, + 0.024445435032248497, + 0.031184131279587746, + 0.034544941037893295, + 0.042515937238931656, + 0.033438790589571, + 0.04217500612139702, + 0.04586270824074745, + 0.05557382106781006, + 0.04323612526059151, + 0.05453914776444435, + 0.00657545356079936 + ], + "innovation_negative_gradient_cosine": [ + 0.9994824528694153, + 0.9993388056755066, + 0.999483585357666, + 0.9993628263473511, + 0.9994879364967346, + 0.9994150996208191, + 0.9994694590568542, + 0.9991943836212158, + 0.9994771480560303, + 0.9993506669998169, + 0.9994744658470154, + 0.9993091821670532, + 0.9994770288467407, + 0.999363899230957, + 0.9994924068450928, + 0.9993765354156494, + 0.9995057582855225, + 0.9994776248931885, + 0.9995115995407104, + 0.9995065927505493, + 0.9995415210723877, + 0.9993911981582642, + 0.9995527267456055, + 0.9994387626647949, + 0.9995670318603516, + 0.9994246363639832, + 0.9995906352996826, + 0.9994567632675171, + 0.999603271484375, + 0.9995082020759583, + 0.9996105432510376, + 0.9995136260986328, + 0.9996271133422852, + 0.9995301365852356, + 0.9996365904808044, + 0.9995220899581909, + 0.9996331930160522, + 0.999711275100708, + 0.999653160572052, + 0.9996668696403503, + 0.9996533989906311, + 0.9996808767318726, + 0.9996672868728638, + 0.9997168779373169, + 0.9997159242630005, + 0.999731183052063, + 0.9996582865715027, + 0.9996111392974854, + 0.9997155666351318, + 0.9997341632843018, + 0.999699056148529, + 0.9996869564056396, + 0.9998605251312256, + 0.9998820424079895, + 0.8115973472595215 + ], + "mean_feedback_forward_cosine": 0.9995472160252658, + "mean_feedback_forward_relative_error": 0.02906279576603662, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9994824528694153, + 0.9993388056755066, + 0.999483585357666, + 0.9993628263473511, + 0.9994879364967346, + 0.9994150996208191, + 0.9994694590568542, + 0.9991943836212158, + 0.9994771480560303, + 0.9993506669998169, + 0.9994744658470154, + 0.9993091821670532, + 0.9994770288467407, + 0.999363899230957, + 0.9994924068450928, + 0.9993765354156494, + 0.9995057582855225, + 0.9994776248931885, + 0.9995115995407104, + 0.9995065927505493, + 0.9995415210723877, + 0.9993911981582642, + 0.9995527267456055, + 0.9994387626647949, + 0.9995670318603516, + 0.9994246363639832, + 0.9995906352996826, + 0.9994567632675171, + 0.999603271484375, + 0.9995082020759583, + 0.9996105432510376, + 0.9995136260986328, + 0.9996271133422852, + 0.9995301365852356, + 0.9996365904808044, + 0.9995220899581909, + 0.9996331930160522, + 0.999711275100708, + 0.999653160572052, + 0.9996668696403503, + 0.9996533989906311, + 0.9996808767318726, + 0.9996672868728638, + 0.9997168779373169, + 0.9997159242630005, + 0.999731183052063, + 0.9996582865715027, + 0.9996111392974854, + 0.9997155666351318, + 0.9997341632843018, + 0.999699056148529, + 0.9996869564056396, + 0.9998605251312256, + 0.9998820424079895, + 0.8115973472595215 + ], + "teaching_negative_gradient_cosine": [ + 0.9994824528694153, + 0.9993388056755066, + 0.999483585357666, + 0.9993628263473511, + 0.9994879364967346, + 0.9994150996208191, + 0.9994694590568542, + 0.9991943836212158, + 0.9994771480560303, + 0.9993506669998169, + 0.9994744658470154, + 0.9993091821670532, + 0.9994770288467407, + 0.999363899230957, + 0.9994924068450928, + 0.9993765354156494, + 0.9995057582855225, + 0.9994776248931885, + 0.9995115995407104, + 0.9995065927505493, + 0.9995415210723877, + 0.9993911981582642, + 0.9995527267456055, + 0.9994387626647949, + 0.9995670318603516, + 0.9994246363639832, + 0.9995906352996826, + 0.9994567632675171, + 0.999603271484375, + 0.9995082020759583, + 0.9996105432510376, + 0.9995136260986328, + 0.9996271133422852, + 0.9995301365852356, + 0.9996365904808044, + 0.9995220899581909, + 0.9996331930160522, + 0.999711275100708, + 0.999653160572052, + 0.9996668696403503, + 0.9996533989906311, + 0.9996808767318726, + 0.9996672868728638, + 0.9997168779373169, + 0.9997159242630005, + 0.999731183052063, + 0.9996582865715027, + 0.9996111392974854, + 0.9997155666351318, + 0.9997341632843018, + 0.999699056148529, + 0.9996869564056396, + 0.9998605251312256, + 0.9998820424079895, + 0.8115973472595215 + ], + "wall_s": 0.1958451271057129 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3753533363342285, + "mean_feedback_forward_cosine": 0.23858112272891133, + "mean_feedback_forward_relative_error": 1.227240677313371, + "min_feedback_forward_cosine": 0.05976404994726181 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.3715767669677734 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3492971658706665, + "mean_feedback_forward_cosine": 0.2779971172863787, + "mean_feedback_forward_relative_error": 1.1959425774487582, + "min_feedback_forward_cosine": 0.09428754448890686 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.7599987884521484 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3413547277450562, + "mean_feedback_forward_cosine": 0.33180117837407375, + "mean_feedback_forward_relative_error": 1.1510712428526444, + "min_feedback_forward_cosine": 0.10544475167989731 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.3671581301879883 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3315541744232178, + "mean_feedback_forward_cosine": 0.3783432330597531, + "mean_feedback_forward_relative_error": 1.11073296232657, + "min_feedback_forward_cosine": 0.12001358717679977 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.1184045793533326 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3182597160339355, + "mean_feedback_forward_cosine": 0.4222381101413207, + "mean_feedback_forward_relative_error": 1.0710616138848392, + "min_feedback_forward_cosine": 0.13862502574920654 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.9559098788452148 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.302269458770752, + "mean_feedback_forward_cosine": 0.46252228292551906, + "mean_feedback_forward_relative_error": 1.0328783571720124, + "min_feedback_forward_cosine": 0.16006067395210266 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8277665467453003 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2851325273513794, + "mean_feedback_forward_cosine": 0.49840348931876094, + "mean_feedback_forward_relative_error": 0.9977190592072227, + "min_feedback_forward_cosine": 0.18014489114284515 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7322896691513061 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2729781866073608, + "mean_feedback_forward_cosine": 0.5305414419282567, + "mean_feedback_forward_relative_error": 0.9645393799651752, + "min_feedback_forward_cosine": 0.19644924998283386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6643429850769043 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2598854303359985, + "mean_feedback_forward_cosine": 0.5591073819182136, + "mean_feedback_forward_relative_error": 0.9337524311109022, + "min_feedback_forward_cosine": 0.21356301009655 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6060961878585815 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2453932762145996, + "mean_feedback_forward_cosine": 0.5866251238367775, + "mean_feedback_forward_relative_error": 0.9031165372241627, + "min_feedback_forward_cosine": 0.23207230865955353 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.56962231924057 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2283390760421753, + "mean_feedback_forward_cosine": 0.6119551198049025, + "mean_feedback_forward_relative_error": 0.8740448279814287, + "min_feedback_forward_cosine": 0.25264260172843933 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5373372936439514 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2145673036575317, + "mean_feedback_forward_cosine": 0.6354925567453558, + "mean_feedback_forward_relative_error": 0.846172617782246, + "min_feedback_forward_cosine": 0.2757928967475891 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5023022323799133 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.200329065322876, + "mean_feedback_forward_cosine": 0.6576699701222506, + "mean_feedback_forward_relative_error": 0.8189129428430038, + "min_feedback_forward_cosine": 0.297935426235199 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.480546227684021 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1844172477722168, + "mean_feedback_forward_cosine": 0.6783240502530878, + "mean_feedback_forward_relative_error": 0.7928331624377858, + "min_feedback_forward_cosine": 0.3203599452972412 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4619817748260498 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1689317226409912, + "mean_feedback_forward_cosine": 0.6979443685574965, + "mean_feedback_forward_relative_error": 0.7671210226687518, + "min_feedback_forward_cosine": 0.3392482101917267 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.43396106838226317 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.152108073234558, + "mean_feedback_forward_cosine": 0.7162653863430023, + "mean_feedback_forward_relative_error": 0.7422217447649349, + "min_feedback_forward_cosine": 0.3589640259742737 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4181311864089966 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1348786354064941, + "mean_feedback_forward_cosine": 0.7338095366954803, + "mean_feedback_forward_relative_error": 0.7178020859306509, + "min_feedback_forward_cosine": 0.37888485193252563 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.40221583515167236 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1163133382797241, + "mean_feedback_forward_cosine": 0.7505907714366913, + "mean_feedback_forward_relative_error": 0.6936617954210802, + "min_feedback_forward_cosine": 0.39955729246139526 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.38683955910682677 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.096354603767395, + "mean_feedback_forward_cosine": 0.7664685032584451, + "mean_feedback_forward_relative_error": 0.6701816315000707, + "min_feedback_forward_cosine": 0.4216799736022949 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.37687294452667236 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.075522780418396, + "mean_feedback_forward_cosine": 0.7813193516297774, + "mean_feedback_forward_relative_error": 0.6474809026176279, + "min_feedback_forward_cosine": 0.44336777925491333 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.35998005069732664 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0544183254241943, + "mean_feedback_forward_cosine": 0.7955406438220631, + "mean_feedback_forward_relative_error": 0.6251224211671136, + "min_feedback_forward_cosine": 0.46595484018325806 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.35017641264915467 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.033109426498413, + "mean_feedback_forward_cosine": 0.808514407006177, + "mean_feedback_forward_relative_error": 0.6041197836399078, + "min_feedback_forward_cosine": 0.4874805510044098 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.33736838609695435 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0119140148162842, + "mean_feedback_forward_cosine": 0.8209045312621377, + "mean_feedback_forward_relative_error": 0.583318842812018, + "min_feedback_forward_cosine": 0.5093268156051636 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3264688785266876 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9892695546150208, + "mean_feedback_forward_cosine": 0.8328157869252292, + "mean_feedback_forward_relative_error": 0.5627880638295953, + "min_feedback_forward_cosine": 0.5313945412635803 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.32062696184158324 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9671341180801392, + "mean_feedback_forward_cosine": 0.8439171400937167, + "mean_feedback_forward_relative_error": 0.5429453088478609, + "min_feedback_forward_cosine": 0.5521942973136902 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31335874647140505 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9442291259765625, + "mean_feedback_forward_cosine": 0.854020869731903, + "mean_feedback_forward_relative_error": 0.5243836768648841, + "min_feedback_forward_cosine": 0.5734392404556274 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2967264782428741 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9206990003585815, + "mean_feedback_forward_cosine": 0.8638439904559743, + "mean_feedback_forward_relative_error": 0.5056991010904313, + "min_feedback_forward_cosine": 0.5943516492843628 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29112177519798277 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8963457942008972, + "mean_feedback_forward_cosine": 0.873080080205744, + "mean_feedback_forward_relative_error": 0.4876327503811229, + "min_feedback_forward_cosine": 0.615612268447876 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2879420889377594 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8727693557739258, + "mean_feedback_forward_cosine": 0.8818722540682012, + "mean_feedback_forward_relative_error": 0.46990829543633894, + "min_feedback_forward_cosine": 0.6358089447021484 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28146713822364805 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8501355648040771, + "mean_feedback_forward_cosine": 0.8899538072672757, + "mean_feedback_forward_relative_error": 0.45294053459709344, + "min_feedback_forward_cosine": 0.654342770576477 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2742295644712448 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8255464434623718, + "mean_feedback_forward_cosine": 0.8975248510187323, + "mean_feedback_forward_relative_error": 0.436677206104452, + "min_feedback_forward_cosine": 0.6740566492080688 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26588446420669554 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8018335700035095, + "mean_feedback_forward_cosine": 0.9046640190211209, + "mean_feedback_forward_relative_error": 0.42078411958434364, + "min_feedback_forward_cosine": 0.6920709609985352 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2695349159812927 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7783890962600708, + "mean_feedback_forward_cosine": 0.911278574033217, + "mean_feedback_forward_relative_error": 0.4056242719292641, + "min_feedback_forward_cosine": 0.7097922563552856 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.25668339400291446 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7556774616241455, + "mean_feedback_forward_cosine": 0.9173880251971158, + "mean_feedback_forward_relative_error": 0.3910275107080286, + "min_feedback_forward_cosine": 0.7264081239700317 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2529852655506134 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7331120371818542, + "mean_feedback_forward_cosine": 0.9231266433542425, + "mean_feedback_forward_relative_error": 0.3769001170992851, + "min_feedback_forward_cosine": 0.7425932288169861 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.24265406985282897 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7092484831809998, + "mean_feedback_forward_cosine": 0.9287206346338446, + "mean_feedback_forward_relative_error": 0.36265066442164506, + "min_feedback_forward_cosine": 0.7588396072387695 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24575304817199706 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.686872661113739, + "mean_feedback_forward_cosine": 0.933739878914573, + "mean_feedback_forward_relative_error": 0.34935173134912145, + "min_feedback_forward_cosine": 0.7736266851425171 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2325409859275818 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6640751361846924, + "mean_feedback_forward_cosine": 0.9385467746041037, + "mean_feedback_forward_relative_error": 0.3362755129283125, + "min_feedback_forward_cosine": 0.7882757186889648 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2368836960887909 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.642682671546936, + "mean_feedback_forward_cosine": 0.9429273377765309, + "mean_feedback_forward_relative_error": 0.3238413172689351, + "min_feedback_forward_cosine": 0.8015169501304626 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.22704051488876342 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6210477352142334, + "mean_feedback_forward_cosine": 0.9470651431517167, + "mean_feedback_forward_relative_error": 0.31177527890963985, + "min_feedback_forward_cosine": 0.8147600889205933 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2250134210729599 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5990719199180603, + "mean_feedback_forward_cosine": 0.9510523882779208, + "mean_feedback_forward_relative_error": 0.2996232462200252, + "min_feedback_forward_cosine": 0.8275330662727356 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22723024928092955 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5792933106422424, + "mean_feedback_forward_cosine": 0.9545747236772018, + "mean_feedback_forward_relative_error": 0.2884721588004719, + "min_feedback_forward_cosine": 0.8386154770851135 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2188000938987732 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5588022470474243, + "mean_feedback_forward_cosine": 0.9579399780793624, + "mean_feedback_forward_relative_error": 0.27745696793903, + "min_feedback_forward_cosine": 0.8496239185333252 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.21868580892562867 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.538779079914093, + "mean_feedback_forward_cosine": 0.9610339793291959, + "mean_feedback_forward_relative_error": 0.26698302951726044, + "min_feedback_forward_cosine": 0.8600154519081116 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.21327681936264037 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5196689367294312, + "mean_feedback_forward_cosine": 0.9639250018379905, + "mean_feedback_forward_relative_error": 0.2568584924394434, + "min_feedback_forward_cosine": 0.8696984052658081 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2090343994283676 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5012470483779907, + "mean_feedback_forward_cosine": 0.966585939580744, + "mean_feedback_forward_relative_error": 0.24716081849553367, + "min_feedback_forward_cosine": 0.8787448406219482 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.20594262621879578 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48256897926330566, + "mean_feedback_forward_cosine": 0.9691115661100907, + "mean_feedback_forward_relative_error": 0.23761470968073065, + "min_feedback_forward_cosine": 0.8873358964920044 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2056799218273163 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.465013325214386, + "mean_feedback_forward_cosine": 0.9714327422055331, + "mean_feedback_forward_relative_error": 0.2285067179663615, + "min_feedback_forward_cosine": 0.8953490853309631 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20313076534509658 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4464685022830963, + "mean_feedback_forward_cosine": 0.9736069094051014, + "mean_feedback_forward_relative_error": 0.2196408591487191, + "min_feedback_forward_cosine": 0.9034064412117004 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20379007936239243 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42973530292510986, + "mean_feedback_forward_cosine": 0.975603437423706, + "mean_feedback_forward_relative_error": 0.2111600256778977, + "min_feedback_forward_cosine": 0.910393238067627 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.1951459578704834 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4128367006778717, + "mean_feedback_forward_cosine": 0.9774770899252458, + "mean_feedback_forward_relative_error": 0.2028905689716339, + "min_feedback_forward_cosine": 0.9172727465629578 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.1930615305519104 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.39636051654815674, + "mean_feedback_forward_cosine": 0.9791891726580533, + "mean_feedback_forward_relative_error": 0.1950628221712329, + "min_feedback_forward_cosine": 0.923702597618103 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.1915807810497284 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38107752799987793, + "mean_feedback_forward_cosine": 0.9807723077860746, + "mean_feedback_forward_relative_error": 0.1875022035430778, + "min_feedback_forward_cosine": 0.9294193983078003 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.1889551772403717 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3662196397781372, + "mean_feedback_forward_cosine": 0.9822606726126237, + "mean_feedback_forward_relative_error": 0.18007641305977648, + "min_feedback_forward_cosine": 0.9347590208053589 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19030168395996094 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.352567583322525, + "mean_feedback_forward_cosine": 0.9835671229795976, + "mean_feedback_forward_relative_error": 0.17330904772335834, + "min_feedback_forward_cosine": 0.939473032951355 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.18390124263048171 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3388565480709076, + "mean_feedback_forward_cosine": 0.9848230567845431, + "mean_feedback_forward_relative_error": 0.1665922653268684, + "min_feedback_forward_cosine": 0.9440066814422607 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.18452228686332703 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3252885341644287, + "mean_feedback_forward_cosine": 0.9859845107251948, + "mean_feedback_forward_relative_error": 0.1601314303549853, + "min_feedback_forward_cosine": 0.9483773708343506 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.18178530210494995 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3123828172683716, + "mean_feedback_forward_cosine": 0.9870757471431385, + "mean_feedback_forward_relative_error": 0.15379290289499542, + "min_feedback_forward_cosine": 0.9523506760597229 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.183211806974411 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30033665895462036, + "mean_feedback_forward_cosine": 0.988059955293482, + "mean_feedback_forward_relative_error": 0.14786873784932222, + "min_feedback_forward_cosine": 0.9559201002120972 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.1828134522819519 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2893706262111664, + "mean_feedback_forward_cosine": 0.9889495286074552, + "mean_feedback_forward_relative_error": 0.14224021170626988, + "min_feedback_forward_cosine": 0.9590567350387573 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.17575358436584473 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2780655026435852, + "mean_feedback_forward_cosine": 0.9897775075652383, + "mean_feedback_forward_relative_error": 0.1368156373500824, + "min_feedback_forward_cosine": 0.9621654748916626 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.17398421032905578 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2671937048435211, + "mean_feedback_forward_cosine": 0.9905405618927695, + "mean_feedback_forward_relative_error": 0.1316248351877386, + "min_feedback_forward_cosine": 0.9650338292121887 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.1702138798713684 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2561606764793396, + "mean_feedback_forward_cosine": 0.9912769718603655, + "mean_feedback_forward_relative_error": 0.1264395376498049, + "min_feedback_forward_cosine": 0.9678294062614441 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.17647948033809663 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24617137014865875, + "mean_feedback_forward_cosine": 0.9919285492463545, + "mean_feedback_forward_relative_error": 0.12165466289628636, + "min_feedback_forward_cosine": 0.970259428024292 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17145303641319276 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23637861013412476, + "mean_feedback_forward_cosine": 0.9925379102880304, + "mean_feedback_forward_relative_error": 0.11700458303093911, + "min_feedback_forward_cosine": 0.9725724458694458 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.16560191806793212 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22676199674606323, + "mean_feedback_forward_cosine": 0.9931233644485473, + "mean_feedback_forward_relative_error": 0.11233989095146006, + "min_feedback_forward_cosine": 0.9747302532196045 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.16914532986164094 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21755750477313995, + "mean_feedback_forward_cosine": 0.9936565973541953, + "mean_feedback_forward_relative_error": 0.1079208682206544, + "min_feedback_forward_cosine": 0.9767346382141113 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.16609831602573394 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20901095867156982, + "mean_feedback_forward_cosine": 0.9941426092928106, + "mean_feedback_forward_relative_error": 0.10371322702955116, + "min_feedback_forward_cosine": 0.9785107374191284 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1675348688173294 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20092988014221191, + "mean_feedback_forward_cosine": 0.994590209830891, + "mean_feedback_forward_relative_error": 0.09967515681954947, + "min_feedback_forward_cosine": 0.9801209568977356 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1662090193605423 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19294902682304382, + "mean_feedback_forward_cosine": 0.9950010332194241, + "mean_feedback_forward_relative_error": 0.09581900222057646, + "min_feedback_forward_cosine": 0.981660008430481 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16575380907058715 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18515849113464355, + "mean_feedback_forward_cosine": 0.9953876462849703, + "mean_feedback_forward_relative_error": 0.09206989001144063, + "min_feedback_forward_cosine": 0.9830981492996216 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.16140517864227294 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17758528888225555, + "mean_feedback_forward_cosine": 0.9957432486794212, + "mean_feedback_forward_relative_error": 0.08848337564956057, + "min_feedback_forward_cosine": 0.9844441413879395 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1637605294418335 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1703767478466034, + "mean_feedback_forward_cosine": 0.9960744359276511, + "mean_feedback_forward_relative_error": 0.08500558852472088, + "min_feedback_forward_cosine": 0.9856750965118408 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16280171936035157 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16341739892959595, + "mean_feedback_forward_cosine": 0.996375438300046, + "mean_feedback_forward_relative_error": 0.08170108446343378, + "min_feedback_forward_cosine": 0.9868167638778687 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16216033221244813 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1568925380706787, + "mean_feedback_forward_cosine": 0.9966534831307151, + "mean_feedback_forward_relative_error": 0.07852334237911485, + "min_feedback_forward_cosine": 0.9878414869308472 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1593558922433853 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15079909563064575, + "mean_feedback_forward_cosine": 0.99690823988481, + "mean_feedback_forward_relative_error": 0.07549517093734308, + "min_feedback_forward_cosine": 0.988762378692627 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.15664530474185945 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14507438242435455, + "mean_feedback_forward_cosine": 0.9971384102647954, + "mean_feedback_forward_relative_error": 0.07264490733769807, + "min_feedback_forward_cosine": 0.9895980358123779 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15697357017040253 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13921557366847992, + "mean_feedback_forward_cosine": 0.9973604917526245, + "mean_feedback_forward_relative_error": 0.06978270492770455, + "min_feedback_forward_cosine": 0.990416944026947 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1557736623430252 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1335224062204361, + "mean_feedback_forward_cosine": 0.9975588289174167, + "mean_feedback_forward_relative_error": 0.0671323896470395, + "min_feedback_forward_cosine": 0.9911783337593079 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15450497795581816 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12787528336048126, + "mean_feedback_forward_cosine": 0.9977456851439043, + "mean_feedback_forward_relative_error": 0.06453609778122468, + "min_feedback_forward_cosine": 0.9919034242630005 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15260362037658692 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12294699251651764, + "mean_feedback_forward_cosine": 0.9979152842001482, + "mean_feedback_forward_relative_error": 0.06206949288872155, + "min_feedback_forward_cosine": 0.9925139546394348 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.15245496213912965 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11825887858867645, + "mean_feedback_forward_cosine": 0.9980723045089028, + "mean_feedback_forward_relative_error": 0.05969261093573137, + "min_feedback_forward_cosine": 0.9930729866027832 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.14681518635749816 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11360993981361389, + "mean_feedback_forward_cosine": 0.9982220443812284, + "mean_feedback_forward_relative_error": 0.05734120180661028, + "min_feedback_forward_cosine": 0.9936062097549438 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15030074099302293 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10911266505718231, + "mean_feedback_forward_cosine": 0.998359325799075, + "mean_feedback_forward_relative_error": 0.055096853101118046, + "min_feedback_forward_cosine": 0.9941010475158691 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1492630126428604 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10453247278928757, + "mean_feedback_forward_cosine": 0.9984864820133555, + "mean_feedback_forward_relative_error": 0.05293607467954809, + "min_feedback_forward_cosine": 0.9945822954177856 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15211886331558228 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10039085149765015, + "mean_feedback_forward_cosine": 0.9986011526801369, + "mean_feedback_forward_relative_error": 0.05090133160014044, + "min_feedback_forward_cosine": 0.995002031326294 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.14970635938167573 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09618359804153442, + "mean_feedback_forward_cosine": 0.9987105098637667, + "mean_feedback_forward_relative_error": 0.04889079356058077, + "min_feedback_forward_cosine": 0.9954111576080322 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15076603876113892 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09216465801000595, + "mean_feedback_forward_cosine": 0.9988103324716742, + "mean_feedback_forward_relative_error": 0.04697381913323294, + "min_feedback_forward_cosine": 0.9957854747772217 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14972564540863037 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08837541192770004, + "mean_feedback_forward_cosine": 0.9989030296152288, + "mean_feedback_forward_relative_error": 0.045120776749470015, + "min_feedback_forward_cosine": 0.9961245059967041 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.15021497856140137 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08487039059400558, + "mean_feedback_forward_cosine": 0.9989857966249639, + "mean_feedback_forward_relative_error": 0.04339315025982532, + "min_feedback_forward_cosine": 0.9964252710342407 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.14488318905830383 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08149995654821396, + "mean_feedback_forward_cosine": 0.9990638689561324, + "mean_feedback_forward_relative_error": 0.041694756072353234, + "min_feedback_forward_cosine": 0.9967029094696045 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14355681781768798 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07798758894205093, + "mean_feedback_forward_cosine": 0.9991379531947049, + "mean_feedback_forward_relative_error": 0.040026017481630496, + "min_feedback_forward_cosine": 0.996979832649231 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1520788672018051 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07485543191432953, + "mean_feedback_forward_cosine": 0.9992034706202421, + "mean_feedback_forward_relative_error": 0.03848231861537153, + "min_feedback_forward_cosine": 0.997217059135437 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1431914198923111 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07189079374074936, + "mean_feedback_forward_cosine": 0.9992649891159752, + "mean_feedback_forward_relative_error": 0.036976625448600815, + "min_feedback_forward_cosine": 0.9974318146705627 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14487737963199615 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06921626627445221, + "mean_feedback_forward_cosine": 0.999318236654455, + "mean_feedback_forward_relative_error": 0.035616513290865855, + "min_feedback_forward_cosine": 0.9976187944412231 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1393111728811264 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06634382903575897, + "mean_feedback_forward_cosine": 0.9993712479417974, + "mean_feedback_forward_relative_error": 0.03421463214538314, + "min_feedback_forward_cosine": 0.9978121519088745 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14071884412527083 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0636911615729332, + "mean_feedback_forward_cosine": 0.9994187701832165, + "mean_feedback_forward_relative_error": 0.03290359213609587, + "min_feedback_forward_cosine": 0.9979828596115112 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.13622581929683686 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06113015115261078, + "mean_feedback_forward_cosine": 0.9994621732018211, + "mean_feedback_forward_relative_error": 0.03166139036078345, + "min_feedback_forward_cosine": 0.9981414079666138 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.13988637959480285 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05856306850910187, + "mean_feedback_forward_cosine": 0.9995021874254401, + "mean_feedback_forward_relative_error": 0.030473380942236294, + "min_feedback_forward_cosine": 0.9982942342758179 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1389887262105942 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05612877383828163, + "mean_feedback_forward_cosine": 0.9995406811887567, + "mean_feedback_forward_relative_error": 0.029282100541009146, + "min_feedback_forward_cosine": 0.9984330534934998 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1357506630563736 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056110598146915436, + "mean_feedback_forward_cosine": 0.9995409900491888, + "mean_feedback_forward_relative_error": 0.029271923581307584, + "min_feedback_forward_cosine": 0.9984340071678162 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07303508335351944 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05610087141394615, + "mean_feedback_forward_cosine": 0.999541123346849, + "mean_feedback_forward_relative_error": 0.029266326561231504, + "min_feedback_forward_cosine": 0.9984345436096191 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05084166835069656 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05609184876084328, + "mean_feedback_forward_cosine": 0.9995412588119507, + "mean_feedback_forward_relative_error": 0.029261213820427658, + "min_feedback_forward_cosine": 0.9984351396560669 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04009084811389446 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05608273670077324, + "mean_feedback_forward_cosine": 0.999541396444494, + "mean_feedback_forward_relative_error": 0.02925638123838739, + "min_feedback_forward_cosine": 0.9984355568885803 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.034704581493735315 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0560733936727047, + "mean_feedback_forward_cosine": 0.9995415373282, + "mean_feedback_forward_relative_error": 0.02925169115716761, + "min_feedback_forward_cosine": 0.9984361529350281 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.030660836898684502 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056064337491989136, + "mean_feedback_forward_cosine": 0.9995416316119108, + "mean_feedback_forward_relative_error": 0.029247118439525367, + "min_feedback_forward_cosine": 0.9984366297721863 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.027093489158153534 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05605465546250343, + "mean_feedback_forward_cosine": 0.9995417800816623, + "mean_feedback_forward_relative_error": 0.02924259211021391, + "min_feedback_forward_cosine": 0.9984371662139893 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.025626796327233314 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05604396015405655, + "mean_feedback_forward_cosine": 0.99954191012816, + "mean_feedback_forward_relative_error": 0.029237881903959945, + "min_feedback_forward_cosine": 0.9984377026557922 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.024102603558897973 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05603327602148056, + "mean_feedback_forward_cosine": 0.9995420585979115, + "mean_feedback_forward_relative_error": 0.029233268165791578, + "min_feedback_forward_cosine": 0.9984383583068848 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.021994172270596028 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05602308362722397, + "mean_feedback_forward_cosine": 0.9995421821420843, + "mean_feedback_forward_relative_error": 0.029228808586909013, + "min_feedback_forward_cosine": 0.9984388947486877 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.02064167999714613 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056012753397226334, + "mean_feedback_forward_cosine": 0.9995423143560236, + "mean_feedback_forward_relative_error": 0.029224336858500132, + "min_feedback_forward_cosine": 0.9984394907951355 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.020304233653247357 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05600287392735481, + "mean_feedback_forward_cosine": 0.9995424368164756, + "mean_feedback_forward_relative_error": 0.029220171382820063, + "min_feedback_forward_cosine": 0.9984400868415833 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01767860997915268 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055992644280195236, + "mean_feedback_forward_cosine": 0.9995425560257651, + "mean_feedback_forward_relative_error": 0.029216162453998218, + "min_feedback_forward_cosine": 0.998440682888031 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.015958202982246875 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05598311126232147, + "mean_feedback_forward_cosine": 0.9995426719838922, + "mean_feedback_forward_relative_error": 0.029212310834025795, + "min_feedback_forward_cosine": 0.998441219329834 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.014467060579061507 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055972620844841, + "mean_feedback_forward_cosine": 0.9995427922769027, + "mean_feedback_forward_relative_error": 0.02920824975113977, + "min_feedback_forward_cosine": 0.998441755771637 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.01501357210099697 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055961620062589645, + "mean_feedback_forward_cosine": 0.9995429212396795, + "mean_feedback_forward_relative_error": 0.029203972245820543, + "min_feedback_forward_cosine": 0.9984422922134399 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.01460032881140709 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05595085769891739, + "mean_feedback_forward_cosine": 0.9995430317792026, + "mean_feedback_forward_relative_error": 0.029199915938079357, + "min_feedback_forward_cosine": 0.9984430074691772 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.013433075107038022 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055939629673957825, + "mean_feedback_forward_cosine": 0.999543145569888, + "mean_feedback_forward_relative_error": 0.02919578467580405, + "min_feedback_forward_cosine": 0.998443603515625 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.013182361350655555 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05592821165919304, + "mean_feedback_forward_cosine": 0.9995432810349898, + "mean_feedback_forward_relative_error": 0.029191561229526996, + "min_feedback_forward_cosine": 0.998444139957428 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.013328653957396746 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05591729283332825, + "mean_feedback_forward_cosine": 0.999543407830325, + "mean_feedback_forward_relative_error": 0.029187543444674124, + "min_feedback_forward_cosine": 0.998444676399231 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.012116881666332483 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05590762570500374, + "mean_feedback_forward_cosine": 0.9995435259558938, + "mean_feedback_forward_relative_error": 0.029183953966606748, + "min_feedback_forward_cosine": 0.9984453916549683 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.010798799199536442 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05589691922068596, + "mean_feedback_forward_cosine": 0.9995436321605335, + "mean_feedback_forward_relative_error": 0.029179944580590183, + "min_feedback_forward_cosine": 0.9984459280967712 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.011075613817870616 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055886756628751755, + "mean_feedback_forward_cosine": 0.9995437318628485, + "mean_feedback_forward_relative_error": 0.029176139501346784, + "min_feedback_forward_cosine": 0.9984466433525085 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.010073255243003369 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055876828730106354, + "mean_feedback_forward_cosine": 0.9995438564907421, + "mean_feedback_forward_relative_error": 0.029172447289932856, + "min_feedback_forward_cosine": 0.9984471797943115 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.009847601467296482 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05586809292435646, + "mean_feedback_forward_cosine": 0.9995439496907321, + "mean_feedback_forward_relative_error": 0.029169122959402473, + "min_feedback_forward_cosine": 0.9984475374221802 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.008988191752880811 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05585648864507675, + "mean_feedback_forward_cosine": 0.9995440851558338, + "mean_feedback_forward_relative_error": 0.029165107350457797, + "min_feedback_forward_cosine": 0.9984482526779175 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.008952195992469788 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05584488436579704, + "mean_feedback_forward_cosine": 0.9995442032814026, + "mean_feedback_forward_relative_error": 0.029161272807554765, + "min_feedback_forward_cosine": 0.9984488487243652 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.008961161298081279 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05583440884947777, + "mean_feedback_forward_cosine": 0.9995443257418546, + "mean_feedback_forward_relative_error": 0.029157466780055652, + "min_feedback_forward_cosine": 0.998449444770813 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.008415414428785442 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05582287907600403, + "mean_feedback_forward_cosine": 0.9995444547046315, + "mean_feedback_forward_relative_error": 0.029153058406981555, + "min_feedback_forward_cosine": 0.9984500408172607 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.009376009065210819 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05581072345376015, + "mean_feedback_forward_cosine": 0.9995445782488043, + "mean_feedback_forward_relative_error": 0.0291485387662595, + "min_feedback_forward_cosine": 0.9984506964683533 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.00964955476462841 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05579955875873566, + "mean_feedback_forward_cosine": 0.9995447158813476, + "mean_feedback_forward_relative_error": 0.02914439800449393, + "min_feedback_forward_cosine": 0.9984513521194458 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.00819142333328724 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05579020828008652, + "mean_feedback_forward_cosine": 0.9995448177511042, + "mean_feedback_forward_relative_error": 0.02914082907478918, + "min_feedback_forward_cosine": 0.998451828956604 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007724165812730789 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055780064314603806, + "mean_feedback_forward_cosine": 0.9995449196208607, + "mean_feedback_forward_relative_error": 0.029137031750922852, + "min_feedback_forward_cosine": 0.9984524846076965 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.007425479439049959 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055770739912986755, + "mean_feedback_forward_cosine": 0.9995450550859625, + "mean_feedback_forward_relative_error": 0.029133377884599295, + "min_feedback_forward_cosine": 0.9984529614448547 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.006916446788460017 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05575976148247719, + "mean_feedback_forward_cosine": 0.9995451580394398, + "mean_feedback_forward_relative_error": 0.029129705451090226, + "min_feedback_forward_cosine": 0.9984534978866577 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.006984633325561881 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05575165897607803, + "mean_feedback_forward_cosine": 0.9995452599091963, + "mean_feedback_forward_relative_error": 0.029126468952745198, + "min_feedback_forward_cosine": 0.9984540343284607 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.006382851370871067 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055739447474479675, + "mean_feedback_forward_cosine": 0.9995453964580189, + "mean_feedback_forward_relative_error": 0.02912215185774998, + "min_feedback_forward_cosine": 0.9984546303749084 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.007196955539733172 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055727098137140274, + "mean_feedback_forward_cosine": 0.999545524337075, + "mean_feedback_forward_relative_error": 0.029117773837325247, + "min_feedback_forward_cosine": 0.9984555244445801 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.007402901085168123 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055713534355163574, + "mean_feedback_forward_cosine": 0.9995456760579889, + "mean_feedback_forward_relative_error": 0.029112934812226078, + "min_feedback_forward_cosine": 0.9984561800956726 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.007763749913834036 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05570363998413086, + "mean_feedback_forward_cosine": 0.9995457898486745, + "mean_feedback_forward_relative_error": 0.029109399545599112, + "min_feedback_forward_cosine": 0.998456597328186 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.005878015903830528 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05569057911634445, + "mean_feedback_forward_cosine": 0.9995459350672635, + "mean_feedback_forward_relative_error": 0.029105050074444576, + "min_feedback_forward_cosine": 0.9984574317932129 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.006828229729831219 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05567900836467743, + "mean_feedback_forward_cosine": 0.999546060778878, + "mean_feedback_forward_relative_error": 0.02910095618360422, + "min_feedback_forward_cosine": 0.9984580278396606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.006254217918552458 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05566876754164696, + "mean_feedback_forward_cosine": 0.9995461604811928, + "mean_feedback_forward_relative_error": 0.029097308506342498, + "min_feedback_forward_cosine": 0.998458743095398 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.005695980285033583 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05565546452999115, + "mean_feedback_forward_cosine": 0.999546328457919, + "mean_feedback_forward_relative_error": 0.029092716916718265, + "min_feedback_forward_cosine": 0.9984594583511353 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006483047075793147 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05564580857753754, + "mean_feedback_forward_cosine": 0.9995464227416299, + "mean_feedback_forward_relative_error": 0.02908916412429376, + "min_feedback_forward_cosine": 0.9984599351882935 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.005303248546496033 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055633921176195145, + "mean_feedback_forward_cosine": 0.9995465430346402, + "mean_feedback_forward_relative_error": 0.029084920036521824, + "min_feedback_forward_cosine": 0.9984606504440308 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.00568860914632678 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05562036111950874, + "mean_feedback_forward_cosine": 0.9995467110113664, + "mean_feedback_forward_relative_error": 0.02908001181415536, + "min_feedback_forward_cosine": 0.9984612464904785 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.005940030800290406 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05560300499200821, + "mean_feedback_forward_cosine": 0.9995468909090216, + "mean_feedback_forward_relative_error": 0.02907430969009345, + "min_feedback_forward_cosine": 0.9984622001647949 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006679464721567929 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05558954179286957, + "mean_feedback_forward_cosine": 0.9995470307090065, + "mean_feedback_forward_relative_error": 0.029069276746701112, + "min_feedback_forward_cosine": 0.9984630346298218 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.005969574514850974 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055577341467142105, + "mean_feedback_forward_cosine": 0.9995471824299206, + "mean_feedback_forward_relative_error": 0.029064963478595017, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005214287418797612 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557722598314285, + "mean_feedback_forward_cosine": 0.9995471705089916, + "mean_feedback_forward_relative_error": 0.02906490105458281, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.0047789658159390095 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557713657617569, + "mean_feedback_forward_cosine": 0.9995471672578291, + "mean_feedback_forward_relative_error": 0.029064845920286395, + "min_feedback_forward_cosine": 0.9984636306762695 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004308129639588296 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055577050894498825, + "mean_feedback_forward_cosine": 0.9995471770113165, + "mean_feedback_forward_relative_error": 0.029064794054085557, + "min_feedback_forward_cosine": 0.9984636306762695 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0042297825917229056 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557696521282196, + "mean_feedback_forward_cosine": 0.9995471802624789, + "mean_feedback_forward_relative_error": 0.029064744550057434, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004123410360803828 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557689443230629, + "mean_feedback_forward_cosine": 0.9995472008531744, + "mean_feedback_forward_relative_error": 0.02906469644301317, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004064238927587867 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557681620121002, + "mean_feedback_forward_cosine": 0.9995471889322455, + "mean_feedback_forward_relative_error": 0.029064647286114367, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.0038662421111017465 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557674169540405, + "mean_feedback_forward_cosine": 0.9995471878485246, + "mean_feedback_forward_relative_error": 0.029064595834775404, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.00411278917491436 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557667836546898, + "mean_feedback_forward_cosine": 0.9995471900159663, + "mean_feedback_forward_relative_error": 0.029064552579075097, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0034468719584774227 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557659640908241, + "mean_feedback_forward_cosine": 0.9995471921834078, + "mean_feedback_forward_relative_error": 0.029064508315853096, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0034755243319645524 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557649955153465, + "mean_feedback_forward_cosine": 0.9995471802624789, + "mean_feedback_forward_relative_error": 0.029064458371563393, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.004001413870276883 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557642877101898, + "mean_feedback_forward_cosine": 0.9995471791787581, + "mean_feedback_forward_relative_error": 0.02906441172076897, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0034888430537469684 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055576350539922714, + "mean_feedback_forward_cosine": 0.9995471867648038, + "mean_feedback_forward_relative_error": 0.029064367863942276, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.003523206314407289 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557628720998764, + "mean_feedback_forward_cosine": 0.9995471965182912, + "mean_feedback_forward_relative_error": 0.02906432424417951, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0032350585881620645 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557620897889137, + "mean_feedback_forward_cosine": 0.9995472008531744, + "mean_feedback_forward_relative_error": 0.02906427825377746, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.003561518343985081 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055576130747795105, + "mean_feedback_forward_cosine": 0.9995472019368952, + "mean_feedback_forward_relative_error": 0.02906423352489417, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.003654435398131609 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055576056241989136, + "mean_feedback_forward_cosine": 0.9995471900159663, + "mean_feedback_forward_relative_error": 0.029064187161963094, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0037600355511903764 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055575959384441376, + "mean_feedback_forward_cosine": 0.9995472084392201, + "mean_feedback_forward_relative_error": 0.02906414189968597, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003887376354970038 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557587742805481, + "mean_feedback_forward_cosine": 0.9995471867648038, + "mean_feedback_forward_relative_error": 0.02906409820372408, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.0031937511023879053 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557580664753914, + "mean_feedback_forward_cosine": 0.999547185681083, + "mean_feedback_forward_relative_error": 0.029064055159687997, + "min_feedback_forward_cosine": 0.9984636306762695 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.003321297593843192 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557574704289436, + "mean_feedback_forward_cosine": 0.9995471976020119, + "mean_feedback_forward_relative_error": 0.02906401441855864, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0030844264754094184 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557568371295929, + "mean_feedback_forward_cosine": 0.9995471954345703, + "mean_feedback_forward_relative_error": 0.029063970350067723, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0032436885320395233 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557561665773392, + "mean_feedback_forward_cosine": 0.9995471900159663, + "mean_feedback_forward_relative_error": 0.029063926459374754, + "min_feedback_forward_cosine": 0.9984636306762695 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0031704610666260123 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557556077837944, + "mean_feedback_forward_cosine": 0.9995471943508495, + "mean_feedback_forward_relative_error": 0.029063886302438648, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.003225151681303978 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557548627257347, + "mean_feedback_forward_cosine": 0.9995471954345703, + "mean_feedback_forward_relative_error": 0.02906383995644071, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.003295274883285165 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0555754229426384, + "mean_feedback_forward_cosine": 0.9995472041043368, + "mean_feedback_forward_relative_error": 0.029063797403465617, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.003042548473458737 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557537451386452, + "mean_feedback_forward_cosine": 0.9995472127741033, + "mean_feedback_forward_relative_error": 0.02906376075168902, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0027685807963088154 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557529255747795, + "mean_feedback_forward_cosine": 0.9995471997694536, + "mean_feedback_forward_relative_error": 0.02906371160325679, + "min_feedback_forward_cosine": 0.9984639286994934 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.003699375984184444 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557522550225258, + "mean_feedback_forward_cosine": 0.9995471976020119, + "mean_feedback_forward_relative_error": 0.0290636691857468, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.003018076034467667 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0555751696228981, + "mean_feedback_forward_cosine": 0.9995472116903825, + "mean_feedback_forward_relative_error": 0.02906362803822214, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.002852506909370422 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055575091391801834, + "mean_feedback_forward_cosine": 0.9995472008531744, + "mean_feedback_forward_relative_error": 0.029063586543568155, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.003247974327132106 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557503178715706, + "mean_feedback_forward_cosine": 0.9995471997694536, + "mean_feedback_forward_relative_error": 0.029063544481654057, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.002867137948926538 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055574964731931686, + "mean_feedback_forward_cosine": 0.9995472106066617, + "mean_feedback_forward_relative_error": 0.029063502555205064, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0031107386077940465 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557490140199661, + "mean_feedback_forward_cosine": 0.9995472030206161, + "mean_feedback_forward_relative_error": 0.029063460002229972, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.00310584289137274 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557483434677124, + "mean_feedback_forward_cosine": 0.9995472095229408, + "mean_feedback_forward_relative_error": 0.02906341812658039, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0031988932495750485 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557479336857796, + "mean_feedback_forward_cosine": 0.9995472073554993, + "mean_feedback_forward_relative_error": 0.029063381754200566, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0027497278521955012 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055574726313352585, + "mean_feedback_forward_cosine": 0.9995472268624739, + "mean_feedback_forward_relative_error": 0.029063340310345995, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.003436942494735122 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557466670870781, + "mean_feedback_forward_cosine": 0.9995472171089866, + "mean_feedback_forward_relative_error": 0.02906330324370753, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.002816575135476887 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557459220290184, + "mean_feedback_forward_cosine": 0.9995472084392201, + "mean_feedback_forward_relative_error": 0.02906325902281837, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0032262862982973457 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557452142238617, + "mean_feedback_forward_cosine": 0.9995472116903825, + "mean_feedback_forward_relative_error": 0.029063219179145314, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.002962641342394054 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0555744469165802, + "mean_feedback_forward_cosine": 0.9995472116903825, + "mean_feedback_forward_relative_error": 0.02906317760829221, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.002968811162225902 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055574383586645126, + "mean_feedback_forward_cosine": 0.9995472225275907, + "mean_feedback_forward_relative_error": 0.02906313966959715, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0027009287017956377 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557432025671005, + "mean_feedback_forward_cosine": 0.9995472116903825, + "mean_feedback_forward_relative_error": 0.029063107183372433, + "min_feedback_forward_cosine": 0.9984636902809143 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0027876313353702425 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557425320148468, + "mean_feedback_forward_cosine": 0.9995472116903825, + "mean_feedback_forward_relative_error": 0.029063069049946287, + "min_feedback_forward_cosine": 0.9984639883041382 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0029258847868442535 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557418614625931, + "mean_feedback_forward_cosine": 0.9995471986857327, + "mean_feedback_forward_relative_error": 0.0290630292147398, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0028698310617730023 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557411536574364, + "mean_feedback_forward_cosine": 0.9995472225275907, + "mean_feedback_forward_relative_error": 0.029062989591197533, + "min_feedback_forward_cosine": 0.9984638094902039 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.002892410690151155 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557403340935707, + "mean_feedback_forward_cosine": 0.9995472290299156, + "mean_feedback_forward_relative_error": 0.029062945395708084, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0031911729769781233 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557398870587349, + "mean_feedback_forward_cosine": 0.999547214941545, + "mean_feedback_forward_relative_error": 0.02906290939585729, + "min_feedback_forward_cosine": 0.9984637498855591 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.002494597954954952 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055573929101228714, + "mean_feedback_forward_cosine": 0.9995472095229408, + "mean_feedback_forward_relative_error": 0.029062868646261366, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0028874194427765906 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557387322187424, + "mean_feedback_forward_cosine": 0.9995472203601491, + "mean_feedback_forward_relative_error": 0.029062832316214388, + "min_feedback_forward_cosine": 0.9984639286994934 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.002676534458883107 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05557382106781006, + "mean_feedback_forward_cosine": 0.9995472160252658, + "mean_feedback_forward_relative_error": 0.02906279576603662, + "min_feedback_forward_cosine": 0.9984638690948486 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0026516037723701446 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9298, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3818249603271484, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3318545920, + "peak_memory_reserved_bytes": 4754243584, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0549397468566895, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 18262.86109972, + "train_wall_s": 18256.888276338577 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250433280000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 5010580480000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json new file mode 100644 index 0000000..628169d --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json @@ -0,0 +1,3654 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9995090001159244, + "feedback_forward_cosine": [ + 0.999492347240448, + 0.9993671178817749, + 0.9994820356369019, + 0.9994138479232788, + 0.9995945692062378, + 0.9995646476745605, + 0.9993652105331421, + 0.9993512034416199, + 0.9994674921035767, + 0.9993777275085449, + 0.9995577931404114, + 0.9995392560958862, + 0.9995492696762085, + 0.9994874000549316, + 0.999705970287323, + 0.9996337294578552, + 0.9997380971908569, + 0.9995861649513245, + 0.9996955990791321, + 0.9997649192810059, + 0.9995940923690796, + 0.9995415210723877, + 0.9996059536933899, + 0.9995549917221069, + 0.9996305704116821, + 0.9995378255844116, + 0.9996811151504517, + 0.9995619654655457, + 0.9997225999832153, + 0.9996045827865601, + 0.9996874928474426, + 0.999572217464447, + 0.9996489882469177, + 0.9994925260543823, + 0.9996796250343323, + 0.9995327591896057, + 0.9997528195381165, + 0.9998396635055542, + 0.9998064041137695, + 0.999785304069519, + 0.9998205304145813, + 0.9997357130050659, + 0.999761700630188, + 0.999605655670166, + 0.9997472763061523, + 0.9995254278182983, + 0.9993584752082825, + 0.9989626407623291, + 0.9987921714782715, + 0.9983224868774414, + 0.9982134103775024, + 0.9976545572280884, + 0.9983347058296204, + 0.997771680355072, + 0.9999784231185913 + ], + "feedback_forward_norm_ratio": [ + 1.000089406967163, + 1.0017415285110474, + 0.9985799789428711, + 1.0014532804489136, + 0.9992918968200684, + 1.000859022140503, + 0.9995941519737244, + 1.001286268234253, + 0.9982871413230896, + 1.000485897064209, + 0.9979104399681091, + 1.000905156135559, + 0.9982880353927612, + 1.001328706741333, + 0.9984083771705627, + 1.0006762742996216, + 0.9997565746307373, + 1.0016711950302124, + 0.9993085265159607, + 1.0002381801605225, + 0.9990171790122986, + 1.0010908842086792, + 0.9994428157806396, + 1.0014185905456543, + 0.9996382594108582, + 1.0020818710327148, + 0.9992923140525818, + 1.0016818046569824, + 0.9992808699607849, + 1.0008541345596313, + 0.9991425275802612, + 1.0014350414276123, + 0.9993600249290466, + 1.0013771057128906, + 0.999732494354248, + 1.0014729499816895, + 0.9994292259216309, + 1.000601053237915, + 1.0000132322311401, + 1.0007882118225098, + 0.9997448325157166, + 1.0010770559310913, + 0.9994710087776184, + 1.0012922286987305, + 0.9996737241744995, + 1.0013598203659058, + 0.9986661076545715, + 1.0019320249557495, + 0.9978883266448975, + 1.0027220249176025, + 0.9968799948692322, + 1.0033371448516846, + 0.997822105884552, + 1.0025888681411743, + 1.0013923645019531 + ], + "feedback_forward_relative_error": [ + 0.03186540678143501, + 0.03565174341201782, + 0.03219625726342201, + 0.034293968230485916, + 0.028475996106863022, + 0.029532907530665398, + 0.03562766686081886, + 0.03606780245900154, + 0.03265102952718735, + 0.03528871014714241, + 0.029782500118017197, + 0.030384186655282974, + 0.030047781765460968, + 0.032069332897663116, + 0.02427857369184494, + 0.02707965299487114, + 0.022884933277964592, + 0.028844140470027924, + 0.02467583678662777, + 0.021682925522327423, + 0.02849097177386284, + 0.03031676448881626, + 0.02807055599987507, + 0.029889831319451332, + 0.027179697528481483, + 0.030506731942296028, + 0.02525198645889759, + 0.029669566079974174, + 0.023555569350719452, + 0.028146808966994286, + 0.025006014853715897, + 0.029306400567293167, + 0.026494760066270828, + 0.031911738216876984, + 0.025312673300504684, + 0.030629947781562805, + 0.022243095561861992, + 0.017927750945091248, + 0.01967710815370083, + 0.02074536494910717, + 0.018943214789032936, + 0.023028654977679253, + 0.02183547243475914, + 0.02812952548265457, + 0.022486107423901558, + 0.03086264058947563, + 0.03581830486655235, + 0.04563266783952713, + 0.04914337396621704, + 0.05806482210755348, + 0.05976491793990135, + 0.06868523359298706, + 0.05769040808081627, + 0.06689479947090149, + 0.006721368059515953 + ], + "innovation_negative_gradient_cosine": [ + 0.9995743632316589, + 0.9992213249206543, + 0.9995816946029663, + 0.9994047284126282, + 0.9995815753936768, + 0.9994869828224182, + 0.9995934963226318, + 0.9992918968200684, + 0.9995976090431213, + 0.9993172883987427, + 0.9996025562286377, + 0.9994415044784546, + 0.9996123909950256, + 0.9994906187057495, + 0.9996334314346313, + 0.9995494484901428, + 0.9996424913406372, + 0.9995386004447937, + 0.9996543526649475, + 0.999616265296936, + 0.9996732473373413, + 0.9994845390319824, + 0.9996790885925293, + 0.9995281100273132, + 0.999685525894165, + 0.9995770454406738, + 0.9996961355209351, + 0.9995557069778442, + 0.9997034072875977, + 0.9996285438537598, + 0.9997194409370422, + 0.9996160268783569, + 0.9997323751449585, + 0.9995778203010559, + 0.9997451305389404, + 0.9996391534805298, + 0.9997557401657104, + 0.9997762441635132, + 0.9997713565826416, + 0.9997782707214355, + 0.9997873902320862, + 0.9997929930686951, + 0.9997757077217102, + 0.9997889995574951, + 0.9998126029968262, + 0.9998635053634644, + 0.9997594356536865, + 0.999828577041626, + 0.9996746778488159, + 0.9996174573898315, + 0.9996563792228699, + 0.9995285272598267, + 0.9997157454490662, + 0.9997916221618652, + 0.6614387035369873 + ], + "mean_feedback_forward_cosine": 0.9994391322135925, + "mean_feedback_forward_relative_error": 0.031407567316835576, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9995743632316589, + 0.9992213249206543, + 0.9995816946029663, + 0.9994047284126282, + 0.9995815753936768, + 0.9994869828224182, + 0.9995934963226318, + 0.9992918968200684, + 0.9995976090431213, + 0.9993172883987427, + 0.9996025562286377, + 0.9994415044784546, + 0.9996123909950256, + 0.9994906187057495, + 0.9996334314346313, + 0.9995494484901428, + 0.9996424913406372, + 0.9995386004447937, + 0.9996543526649475, + 0.999616265296936, + 0.9996732473373413, + 0.9994845390319824, + 0.9996790885925293, + 0.9995281100273132, + 0.999685525894165, + 0.9995770454406738, + 0.9996961355209351, + 0.9995557069778442, + 0.9997034072875977, + 0.9996285438537598, + 0.9997194409370422, + 0.9996160268783569, + 0.9997323751449585, + 0.9995778203010559, + 0.9997451305389404, + 0.9996391534805298, + 0.9997557401657104, + 0.9997762441635132, + 0.9997713565826416, + 0.9997782707214355, + 0.9997873902320862, + 0.9997929930686951, + 0.9997757077217102, + 0.9997889995574951, + 0.9998126029968262, + 0.9998635053634644, + 0.9997594356536865, + 0.999828577041626, + 0.9996746778488159, + 0.9996174573898315, + 0.9996563792228699, + 0.9995285272598267, + 0.9997157454490662, + 0.9997916221618652, + 0.6614387035369873 + ], + "teaching_negative_gradient_cosine": [ + 0.9995743632316589, + 0.9992213249206543, + 0.9995816946029663, + 0.9994047284126282, + 0.9995815753936768, + 0.9994869828224182, + 0.9995934963226318, + 0.9992918968200684, + 0.9995976090431213, + 0.9993172883987427, + 0.9996025562286377, + 0.9994415044784546, + 0.9996123909950256, + 0.9994906187057495, + 0.9996334314346313, + 0.9995494484901428, + 0.9996424913406372, + 0.9995386004447937, + 0.9996543526649475, + 0.999616265296936, + 0.9996732473373413, + 0.9994845390319824, + 0.9996790885925293, + 0.9995281100273132, + 0.999685525894165, + 0.9995770454406738, + 0.9996961355209351, + 0.9995557069778442, + 0.9997034072875977, + 0.9996285438537598, + 0.9997194409370422, + 0.9996160268783569, + 0.9997323751449585, + 0.9995778203010559, + 0.9997451305389404, + 0.9996391534805298, + 0.9997557401657104, + 0.9997762441635132, + 0.9997713565826416, + 0.9997782707214355, + 0.9997873902320862, + 0.9997929930686951, + 0.9997757077217102, + 0.9997889995574951, + 0.9998126029968262, + 0.9998635053634644, + 0.9997594356536865, + 0.999828577041626, + 0.9996746778488159, + 0.9996174573898315, + 0.9996563792228699, + 0.9995285272598267, + 0.9997157454490662, + 0.9997916221618652, + 0.6614387035369873 + ], + "wall_s": 0.191314697265625 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3244383335113525, + "mean_feedback_forward_cosine": 0.4198355341499502, + "mean_feedback_forward_relative_error": 1.0560421732338992, + "min_feedback_forward_cosine": 0.127158984541893 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.4138239181900025 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3209409713745117, + "mean_feedback_forward_cosine": 0.4342506834051826, + "mean_feedback_forward_relative_error": 1.0433166081255132, + "min_feedback_forward_cosine": 0.1320267617702484 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.7105659466934204 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3159164190292358, + "mean_feedback_forward_cosine": 0.4584254817529158, + "mean_feedback_forward_relative_error": 1.0215907725420865, + "min_feedback_forward_cosine": 0.13956084847450256 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4826833779144286 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.308376431465149, + "mean_feedback_forward_cosine": 0.4871980583125895, + "mean_feedback_forward_relative_error": 0.9948613367297433, + "min_feedback_forward_cosine": 0.14930999279022217 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.27878708026886 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2998486757278442, + "mean_feedback_forward_cosine": 0.5155361040072007, + "mean_feedback_forward_relative_error": 0.9674677659164775, + "min_feedback_forward_cosine": 0.1607993245124817 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.0934885717010498 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2901713848114014, + "mean_feedback_forward_cosine": 0.5428229207342321, + "mean_feedback_forward_relative_error": 0.9398032789880579, + "min_feedback_forward_cosine": 0.1737784445285797 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.9438204985427856 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2803062200546265, + "mean_feedback_forward_cosine": 0.567431657964533, + "mean_feedback_forward_relative_error": 0.9138331499966708, + "min_feedback_forward_cosine": 0.1865416318178177 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.8270858850860596 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2699190378189087, + "mean_feedback_forward_cosine": 0.5901177953590047, + "mean_feedback_forward_relative_error": 0.8889968286861073, + "min_feedback_forward_cosine": 0.19944104552268982 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.7495757960128784 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2612634897232056, + "mean_feedback_forward_cosine": 0.610681467977437, + "mean_feedback_forward_relative_error": 0.865493153442036, + "min_feedback_forward_cosine": 0.20822472870349884 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.679872678489685 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2559030055999756, + "mean_feedback_forward_cosine": 0.6305192985317923, + "mean_feedback_forward_relative_error": 0.8422314649278467, + "min_feedback_forward_cosine": 0.21694274246692657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.6310647740745544 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2501100301742554, + "mean_feedback_forward_cosine": 0.6483281940221787, + "mean_feedback_forward_relative_error": 0.8205998350273479, + "min_feedback_forward_cosine": 0.22608375549316406 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5876526419639587 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2440547943115234, + "mean_feedback_forward_cosine": 0.6655318002809177, + "mean_feedback_forward_relative_error": 0.7990088679573752, + "min_feedback_forward_cosine": 0.23492339253425598 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5574879420471192 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2371782064437866, + "mean_feedback_forward_cosine": 0.6813188051635569, + "mean_feedback_forward_relative_error": 0.7786574320359664, + "min_feedback_forward_cosine": 0.24500702321529388 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.5174666587638855 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.229864478111267, + "mean_feedback_forward_cosine": 0.6966934740543366, + "mean_feedback_forward_relative_error": 0.7583921773867174, + "min_feedback_forward_cosine": 0.2564243674278259 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4944457879543304 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2204899787902832, + "mean_feedback_forward_cosine": 0.7118308224461295, + "mean_feedback_forward_relative_error": 0.7377489366314628, + "min_feedback_forward_cosine": 0.26965150237083435 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4681345001411438 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.21138334274292, + "mean_feedback_forward_cosine": 0.7262426392598585, + "mean_feedback_forward_relative_error": 0.7175799703056162, + "min_feedback_forward_cosine": 0.28238144516944885 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.45293981085777285 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2019273042678833, + "mean_feedback_forward_cosine": 0.7394660765474493, + "mean_feedback_forward_relative_error": 0.6985175696286288, + "min_feedback_forward_cosine": 0.2950320839881897 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.4310007293605804 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1909077167510986, + "mean_feedback_forward_cosine": 0.7525688263502988, + "mean_feedback_forward_relative_error": 0.6793072627349334, + "min_feedback_forward_cosine": 0.30850374698638916 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.4120278524303436 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1793495416641235, + "mean_feedback_forward_cosine": 0.7650805679234591, + "mean_feedback_forward_relative_error": 0.6604170032522895, + "min_feedback_forward_cosine": 0.3222784996032715 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3957406766605377 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1666817665100098, + "mean_feedback_forward_cosine": 0.7772322437979958, + "mean_feedback_forward_relative_error": 0.6416691671718251, + "min_feedback_forward_cosine": 0.33726298809051514 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3890600772857666 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.15311861038208, + "mean_feedback_forward_cosine": 0.7886456803842025, + "mean_feedback_forward_relative_error": 0.6236146463589235, + "min_feedback_forward_cosine": 0.353469580411911 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.36867258852005 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1378247737884521, + "mean_feedback_forward_cosine": 0.7999685997312719, + "mean_feedback_forward_relative_error": 0.6052621527151628, + "min_feedback_forward_cosine": 0.3707684278488159 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3646633823204041 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.121764898300171, + "mean_feedback_forward_cosine": 0.8108454075726595, + "mean_feedback_forward_relative_error": 0.5872186479243365, + "min_feedback_forward_cosine": 0.38909459114074707 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3533440400314331 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1063214540481567, + "mean_feedback_forward_cosine": 0.821115351265127, + "mean_feedback_forward_relative_error": 0.5697222119027918, + "min_feedback_forward_cosine": 0.40641462802886963 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3383803585338592 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0895359516143799, + "mean_feedback_forward_cosine": 0.8312289080836556, + "mean_feedback_forward_relative_error": 0.552090187235312, + "min_feedback_forward_cosine": 0.42520034313201904 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3309914978790283 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.073138952255249, + "mean_feedback_forward_cosine": 0.8406367025592111, + "mean_feedback_forward_relative_error": 0.5351539647037332, + "min_feedback_forward_cosine": 0.4425792098045349 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.320999254283905 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0543512105941772, + "mean_feedback_forward_cosine": 0.8497189771045338, + "mean_feedback_forward_relative_error": 0.5185014180161737, + "min_feedback_forward_cosine": 0.462225079536438 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.31371587188243866 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0364478826522827, + "mean_feedback_forward_cosine": 0.8583152397112412, + "mean_feedback_forward_relative_error": 0.5023321167989211, + "min_feedback_forward_cosine": 0.48109471797943115 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3024843281364441 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.017153024673462, + "mean_feedback_forward_cosine": 0.8665038390593095, + "mean_feedback_forward_relative_error": 0.48660044859756124, + "min_feedback_forward_cosine": 0.5008649826049805 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2939609637641907 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9980878829956055, + "mean_feedback_forward_cosine": 0.8744109912352128, + "mean_feedback_forward_relative_error": 0.4709355667233467, + "min_feedback_forward_cosine": 0.5197159051895142 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.28885329203605653 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9772093892097473, + "mean_feedback_forward_cosine": 0.8819113514640115, + "mean_feedback_forward_relative_error": 0.4558220600539988, + "min_feedback_forward_cosine": 0.5398478507995605 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2821906175947189 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9556038975715637, + "mean_feedback_forward_cosine": 0.8891913132234053, + "mean_feedback_forward_relative_error": 0.4406657814979553, + "min_feedback_forward_cosine": 0.5601248741149902 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2786395811319351 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9322633147239685, + "mean_feedback_forward_cosine": 0.8961453188549389, + "mean_feedback_forward_relative_error": 0.4259107045151971, + "min_feedback_forward_cosine": 0.5821226239204407 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.27072911348342893 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9096354842185974, + "mean_feedback_forward_cosine": 0.9028071836991743, + "mean_feedback_forward_relative_error": 0.41126837974244895, + "min_feedback_forward_cosine": 0.6022782921791077 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.26704496616363527 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8879315257072449, + "mean_feedback_forward_cosine": 0.9087696888230063, + "mean_feedback_forward_relative_error": 0.39780180914835495, + "min_feedback_forward_cosine": 0.6216888427734375 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25724610758781435 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8641149401664734, + "mean_feedback_forward_cosine": 0.9146937760439786, + "mean_feedback_forward_relative_error": 0.38405355621467935, + "min_feedback_forward_cosine": 0.6411852836608887 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.257418413066864 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.840953528881073, + "mean_feedback_forward_cosine": 0.9202366232872009, + "mean_feedback_forward_relative_error": 0.37086766606027427, + "min_feedback_forward_cosine": 0.6600959300994873 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24879705251693726 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8156360387802124, + "mean_feedback_forward_cosine": 0.9256891088052229, + "mean_feedback_forward_relative_error": 0.3574880524115129, + "min_feedback_forward_cosine": 0.6800718307495117 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24602906025886537 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7926400899887085, + "mean_feedback_forward_cosine": 0.9306280038573526, + "mean_feedback_forward_relative_error": 0.3450457442890514, + "min_feedback_forward_cosine": 0.6978662014007568 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2412323922920227 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7697350978851318, + "mean_feedback_forward_cosine": 0.9352672685276379, + "mean_feedback_forward_relative_error": 0.33294801833954724, + "min_feedback_forward_cosine": 0.7149748206138611 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2362298415184021 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7469183206558228, + "mean_feedback_forward_cosine": 0.9396627296100963, + "mean_feedback_forward_relative_error": 0.32107209251685576, + "min_feedback_forward_cosine": 0.7316209077835083 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23337628873825073 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7224966287612915, + "mean_feedback_forward_cosine": 0.9438528407703747, + "mean_feedback_forward_relative_error": 0.3094679505987601, + "min_feedback_forward_cosine": 0.749053955078125 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23098030502319336 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6987288594245911, + "mean_feedback_forward_cosine": 0.9477465358647433, + "mean_feedback_forward_relative_error": 0.29833831367167557, + "min_feedback_forward_cosine": 0.7652150392532349 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2250031017637253 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6747825145721436, + "mean_feedback_forward_cosine": 0.9515370965003968, + "mean_feedback_forward_relative_error": 0.2871929381381382, + "min_feedback_forward_cosine": 0.7807072401046753 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.2254936946439743 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6516567468643188, + "mean_feedback_forward_cosine": 0.955012834072113, + "mean_feedback_forward_relative_error": 0.2764912331646139, + "min_feedback_forward_cosine": 0.7951079607009888 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22084568507671357 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.628865659236908, + "mean_feedback_forward_cosine": 0.9582318782806396, + "mean_feedback_forward_relative_error": 0.2662767881696874, + "min_feedback_forward_cosine": 0.8090265989303589 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21951218831062316 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6079846620559692, + "mean_feedback_forward_cosine": 0.9611279325051741, + "mean_feedback_forward_relative_error": 0.2566930052908984, + "min_feedback_forward_cosine": 0.8213279247283936 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21334740138053895 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5870092511177063, + "mean_feedback_forward_cosine": 0.9639066804539074, + "mean_feedback_forward_relative_error": 0.24725833399729294, + "min_feedback_forward_cosine": 0.8334741592407227 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2046345087623596 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5655739903450012, + "mean_feedback_forward_cosine": 0.9666054834019054, + "mean_feedback_forward_relative_error": 0.23775543746623126, + "min_feedback_forward_cosine": 0.8452627062797546 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21139758206367493 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5459854602813721, + "mean_feedback_forward_cosine": 0.9689616495912725, + "mean_feedback_forward_relative_error": 0.22913335771723228, + "min_feedback_forward_cosine": 0.8555338978767395 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2025952133321762 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.526841938495636, + "mean_feedback_forward_cosine": 0.9711984493515708, + "mean_feedback_forward_relative_error": 0.2206283477896994, + "min_feedback_forward_cosine": 0.8654561042785645 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20171171308994293 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5071823596954346, + "mean_feedback_forward_cosine": 0.9733563986691561, + "mean_feedback_forward_relative_error": 0.21216394819996573, + "min_feedback_forward_cosine": 0.8751754760742188 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20021353316307067 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4873877167701721, + "mean_feedback_forward_cosine": 0.9753846992145885, + "mean_feedback_forward_relative_error": 0.20394220799207688, + "min_feedback_forward_cosine": 0.8846782445907593 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19994908579826354 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46815577149391174, + "mean_feedback_forward_cosine": 0.9772642558271235, + "mean_feedback_forward_relative_error": 0.19602264843203804, + "min_feedback_forward_cosine": 0.8935234546661377 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19712803175926208 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45061376690864563, + "mean_feedback_forward_cosine": 0.9789202657612888, + "mean_feedback_forward_relative_error": 0.18874934566291895, + "min_feedback_forward_cosine": 0.9012744426727295 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.1897450712442398 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43402156233787537, + "mean_feedback_forward_cosine": 0.9804969830946488, + "mean_feedback_forward_relative_error": 0.1815253614702008, + "min_feedback_forward_cosine": 0.9083393812179565 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1948779441356659 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41747668385505676, + "mean_feedback_forward_cosine": 0.9819663166999817, + "mean_feedback_forward_relative_error": 0.17460938387296415, + "min_feedback_forward_cosine": 0.9150904417037964 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1941714857006073 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4008060693740845, + "mean_feedback_forward_cosine": 0.983362311666662, + "mean_feedback_forward_relative_error": 0.1677873245017095, + "min_feedback_forward_cosine": 0.9216729402542114 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19348405673980712 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3854062855243683, + "mean_feedback_forward_cosine": 0.9845780177549882, + "mean_feedback_forward_relative_error": 0.16157968430356545, + "min_feedback_forward_cosine": 0.9275352954864502 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18416994456768035 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36981651186943054, + "mean_feedback_forward_cosine": 0.9857558510520241, + "mean_feedback_forward_relative_error": 0.15531217523596502, + "min_feedback_forward_cosine": 0.9332382678985596 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18529998634815217 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3562355637550354, + "mean_feedback_forward_cosine": 0.9868420557542281, + "mean_feedback_forward_relative_error": 0.14927968003533104, + "min_feedback_forward_cosine": 0.9379853010177612 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18291772512435914 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3428122401237488, + "mean_feedback_forward_cosine": 0.9878508589484475, + "mean_feedback_forward_relative_error": 0.14348044361580503, + "min_feedback_forward_cosine": 0.942572832107544 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18351415759563447 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32837072014808655, + "mean_feedback_forward_cosine": 0.9888015866279602, + "mean_feedback_forward_relative_error": 0.1378271481530233, + "min_feedback_forward_cosine": 0.9472540020942688 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1803669850063324 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31547480821609497, + "mean_feedback_forward_cosine": 0.9896420522169633, + "mean_feedback_forward_relative_error": 0.13261645124717192, + "min_feedback_forward_cosine": 0.9512698650360107 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17748355750083925 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3037189543247223, + "mean_feedback_forward_cosine": 0.9904208053242076, + "mean_feedback_forward_relative_error": 0.12756685563786463, + "min_feedback_forward_cosine": 0.9548318386077881 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.17406793964862824 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2912445664405823, + "mean_feedback_forward_cosine": 0.9911755648526278, + "mean_feedback_forward_relative_error": 0.12248950502411886, + "min_feedback_forward_cosine": 0.958398699760437 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.17906529505729676 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2801404893398285, + "mean_feedback_forward_cosine": 0.9918416305021807, + "mean_feedback_forward_relative_error": 0.11781480559571222, + "min_feedback_forward_cosine": 0.9614805579185486 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.173457589635849 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27006998658180237, + "mean_feedback_forward_cosine": 0.9924406051635742, + "mean_feedback_forward_relative_error": 0.1134418520737778, + "min_feedback_forward_cosine": 0.9641738533973694 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.16886799769878388 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2597082555294037, + "mean_feedback_forward_cosine": 0.9930091619491577, + "mean_feedback_forward_relative_error": 0.10910528938878666, + "min_feedback_forward_cosine": 0.9668391942977905 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.16880864161491393 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24950578808784485, + "mean_feedback_forward_cosine": 0.9935406565666198, + "mean_feedback_forward_relative_error": 0.10489608872343194, + "min_feedback_forward_cosine": 0.969358503818512 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16825597145080567 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23873963952064514, + "mean_feedback_forward_cosine": 0.9940642356872559, + "mean_feedback_forward_relative_error": 0.10061055208471688, + "min_feedback_forward_cosine": 0.9719176292419434 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1729454252386093 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22891759872436523, + "mean_feedback_forward_cosine": 0.9945134357972578, + "mean_feedback_forward_relative_error": 0.09677447625859217, + "min_feedback_forward_cosine": 0.9741685390472412 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16978507235527038 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21945972740650177, + "mean_feedback_forward_cosine": 0.9949410059235313, + "mean_feedback_forward_relative_error": 0.0929661270908334, + "min_feedback_forward_cosine": 0.9762440919876099 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16630286949157716 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21060435473918915, + "mean_feedback_forward_cosine": 0.9953264767473394, + "mean_feedback_forward_relative_error": 0.08938674618574706, + "min_feedback_forward_cosine": 0.9781172871589661 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16400834189891816 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20204050838947296, + "mean_feedback_forward_cosine": 0.9956956917589361, + "mean_feedback_forward_relative_error": 0.08582023791968822, + "min_feedback_forward_cosine": 0.9798524379730225 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.16743798432826995 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19352245330810547, + "mean_feedback_forward_cosine": 0.9960233428261497, + "mean_feedback_forward_relative_error": 0.08252717652781442, + "min_feedback_forward_cosine": 0.9815119504928589 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1595230025625229 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1849481463432312, + "mean_feedback_forward_cosine": 0.9963419870896773, + "mean_feedback_forward_relative_error": 0.07919774959710511, + "min_feedback_forward_cosine": 0.9830996990203857 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.16649277725219727 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1773487776517868, + "mean_feedback_forward_cosine": 0.9966184778646989, + "mean_feedback_forward_relative_error": 0.07619481232355942, + "min_feedback_forward_cosine": 0.9844546318054199 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15958376306056976 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1702713668346405, + "mean_feedback_forward_cosine": 0.9968827930363742, + "mean_feedback_forward_relative_error": 0.07318445687944239, + "min_feedback_forward_cosine": 0.9856667518615723 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16138536895275116 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16331738233566284, + "mean_feedback_forward_cosine": 0.9971186236901717, + "mean_feedback_forward_relative_error": 0.0704044618728486, + "min_feedback_forward_cosine": 0.9868065118789673 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15965634570121764 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1560080200433731, + "mean_feedback_forward_cosine": 0.9973478187214244, + "mean_feedback_forward_relative_error": 0.06760258132761175, + "min_feedback_forward_cosine": 0.9879544973373413 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.15754901772499086 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14889590442180634, + "mean_feedback_forward_cosine": 0.9975626056844538, + "mean_feedback_forward_relative_error": 0.06485398079861295, + "min_feedback_forward_cosine": 0.9890235662460327 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1604923850774765 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1428852528333664, + "mean_feedback_forward_cosine": 0.9977468035437844, + "mean_feedback_forward_relative_error": 0.06238936851309104, + "min_feedback_forward_cosine": 0.9898895025253296 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15494182635784148 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13744902610778809, + "mean_feedback_forward_cosine": 0.9979130354794589, + "mean_feedback_forward_relative_error": 0.06006247954951091, + "min_feedback_forward_cosine": 0.9906392693519592 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1530601090836525 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13181516528129578, + "mean_feedback_forward_cosine": 0.9980757865038785, + "mean_feedback_forward_relative_error": 0.057702659544619644, + "min_feedback_forward_cosine": 0.991387128829956 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15579026302337645 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1261598914861679, + "mean_feedback_forward_cosine": 0.9982293020595204, + "mean_feedback_forward_relative_error": 0.05538785102015192, + "min_feedback_forward_cosine": 0.9921053647994995 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.15822976639270783 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12078189104795456, + "mean_feedback_forward_cosine": 0.9983660004355691, + "mean_feedback_forward_relative_error": 0.05325097621164539, + "min_feedback_forward_cosine": 0.992760419845581 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15028054527759552 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11607860028743744, + "mean_feedback_forward_cosine": 0.9984901482408697, + "mean_feedback_forward_relative_error": 0.05120365558700128, + "min_feedback_forward_cosine": 0.9933149814605713 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.15129416436910628 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11128923296928406, + "mean_feedback_forward_cosine": 0.9986075195399198, + "mean_feedback_forward_relative_error": 0.049207691560414706, + "min_feedback_forward_cosine": 0.9938527941703796 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.14803420697927475 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10640254616737366, + "mean_feedback_forward_cosine": 0.998720792206851, + "mean_feedback_forward_relative_error": 0.047186382792212746, + "min_feedback_forward_cosine": 0.9943797588348389 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.153843375749588 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10187916457653046, + "mean_feedback_forward_cosine": 0.9988207470286976, + "mean_feedback_forward_relative_error": 0.0453362681648948, + "min_feedback_forward_cosine": 0.994846522808075 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14836268123626709 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09747994691133499, + "mean_feedback_forward_cosine": 0.99891292181882, + "mean_feedback_forward_relative_error": 0.04355126805603504, + "min_feedback_forward_cosine": 0.9952813386917114 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1528516566991806 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09351164847612381, + "mean_feedback_forward_cosine": 0.9989956454797224, + "mean_feedback_forward_relative_error": 0.041882835277779536, + "min_feedback_forward_cosine": 0.9956561326980591 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14722164509296418 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08970483392477036, + "mean_feedback_forward_cosine": 0.9990740808573636, + "mean_feedback_forward_relative_error": 0.0402288471263918, + "min_feedback_forward_cosine": 0.9960024356842041 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14886850304365157 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08588581532239914, + "mean_feedback_forward_cosine": 0.9991466793147, + "mean_feedback_forward_relative_error": 0.03864135615188967, + "min_feedback_forward_cosine": 0.9963352084159851 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.14826063853263854 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223635703325272, + "mean_feedback_forward_cosine": 0.9992124145681208, + "mean_feedback_forward_relative_error": 0.037147075022486126, + "min_feedback_forward_cosine": 0.9966393113136292 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14491665635108947 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0788571760058403, + "mean_feedback_forward_cosine": 0.9992724895477295, + "mean_feedback_forward_relative_error": 0.03572304660285061, + "min_feedback_forward_cosine": 0.9969098567962646 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14346269236564635 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07558047026395798, + "mean_feedback_forward_cosine": 0.9993296547369523, + "mean_feedback_forward_relative_error": 0.034297426011074675, + "min_feedback_forward_cosine": 0.9971612691879272 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.14584156653642655 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07237829267978668, + "mean_feedback_forward_cosine": 0.9993817773732272, + "mean_feedback_forward_relative_error": 0.03295629530806433, + "min_feedback_forward_cosine": 0.9973962903022766 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14407524058818816 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06947639584541321, + "mean_feedback_forward_cosine": 0.999428694898432, + "mean_feedback_forward_relative_error": 0.03169208192689852, + "min_feedback_forward_cosine": 0.9975995421409607 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14411247708320618 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06945092976093292, + "mean_feedback_forward_cosine": 0.999429096958854, + "mean_feedback_forward_relative_error": 0.031680582599206406, + "min_feedback_forward_cosine": 0.997601330280304 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07716866317510605 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06943776458501816, + "mean_feedback_forward_cosine": 0.9994292963634838, + "mean_feedback_forward_relative_error": 0.03167427863756364, + "min_feedback_forward_cosine": 0.9976023435592651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.051544788405299184 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0694250762462616, + "mean_feedback_forward_cosine": 0.9994294773448598, + "mean_feedback_forward_relative_error": 0.03166841171173887, + "min_feedback_forward_cosine": 0.9976032972335815 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04210718454018235 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06941260397434235, + "mean_feedback_forward_cosine": 0.999429668079723, + "mean_feedback_forward_relative_error": 0.031662963059815495, + "min_feedback_forward_cosine": 0.9976043701171875 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03675717756778002 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06940004974603653, + "mean_feedback_forward_cosine": 0.9994298501448198, + "mean_feedback_forward_relative_error": 0.03165755365890535, + "min_feedback_forward_cosine": 0.9976052045822144 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.031186733311414718 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0693858414888382, + "mean_feedback_forward_cosine": 0.9994300560517745, + "mean_feedback_forward_relative_error": 0.031651697697287255, + "min_feedback_forward_cosine": 0.9976061582565308 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.0309412691372633 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06937147676944733, + "mean_feedback_forward_cosine": 0.9994302467866377, + "mean_feedback_forward_relative_error": 0.03164597680284218, + "min_feedback_forward_cosine": 0.9976072311401367 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.028081531404852866 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06935874372720718, + "mean_feedback_forward_cosine": 0.9994304375215011, + "mean_feedback_forward_relative_error": 0.03164071908051318, + "min_feedback_forward_cosine": 0.9976081848144531 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.02442640140235424 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0693446695804596, + "mean_feedback_forward_cosine": 0.999430613084273, + "mean_feedback_forward_relative_error": 0.031635197048837486, + "min_feedback_forward_cosine": 0.99760901927948 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.023568274050354956 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06933095306158066, + "mean_feedback_forward_cosine": 0.9994308016516945, + "mean_feedback_forward_relative_error": 0.031630029351535166, + "min_feedback_forward_cosine": 0.9976098537445068 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.020625956691354513 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06931696087121964, + "mean_feedback_forward_cosine": 0.9994309826330705, + "mean_feedback_forward_relative_error": 0.031624852019277484, + "min_feedback_forward_cosine": 0.9976110458374023 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.020318002663254738 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06930254399776459, + "mean_feedback_forward_cosine": 0.9994311701167714, + "mean_feedback_forward_relative_error": 0.03161953128027645, + "min_feedback_forward_cosine": 0.9976118803024292 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.018820216268151998 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0692884624004364, + "mean_feedback_forward_cosine": 0.9994313651865179, + "mean_feedback_forward_relative_error": 0.0316144339824942, + "min_feedback_forward_cosine": 0.9976129531860352 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.017537187362760307 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06927390396595001, + "mean_feedback_forward_cosine": 0.9994315515864979, + "mean_feedback_forward_relative_error": 0.03160912807692181, + "min_feedback_forward_cosine": 0.9976140856742859 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01681633037008345 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06925929337739944, + "mean_feedback_forward_cosine": 0.9994317444888028, + "mean_feedback_forward_relative_error": 0.03160384031520649, + "min_feedback_forward_cosine": 0.9976149201393127 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.016025505062937737 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06924420595169067, + "mean_feedback_forward_cosine": 0.9994319384748286, + "mean_feedback_forward_relative_error": 0.03159833199937235, + "min_feedback_forward_cosine": 0.9976159334182739 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.01605200676083565 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06923021376132965, + "mean_feedback_forward_cosine": 0.9994321400469, + "mean_feedback_forward_relative_error": 0.03159324789250439, + "min_feedback_forward_cosine": 0.9976168870925903 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.013920427777171136 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06921586394309998, + "mean_feedback_forward_cosine": 0.9994323199445551, + "mean_feedback_forward_relative_error": 0.0315880704247816, + "min_feedback_forward_cosine": 0.9976179599761963 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.013951102841943503 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06920035928487778, + "mean_feedback_forward_cosine": 0.9994325269352305, + "mean_feedback_forward_relative_error": 0.031582749711180275, + "min_feedback_forward_cosine": 0.997619092464447 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.013470773504525423 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06918603926897049, + "mean_feedback_forward_cosine": 0.999432711167769, + "mean_feedback_forward_relative_error": 0.03157752440734343, + "min_feedback_forward_cosine": 0.9976199865341187 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.013048178313672543 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06917116791009903, + "mean_feedback_forward_cosine": 0.999432897567749, + "mean_feedback_forward_relative_error": 0.031572245416993445, + "min_feedback_forward_cosine": 0.9976210594177246 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.011964423420876264 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06915569305419922, + "mean_feedback_forward_cosine": 0.9994331078095869, + "mean_feedback_forward_relative_error": 0.03156682039526376, + "min_feedback_forward_cosine": 0.9976219534873962 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.012266639451049268 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06913956254720688, + "mean_feedback_forward_cosine": 0.9994333267211915, + "mean_feedback_forward_relative_error": 0.03156118229539557, + "min_feedback_forward_cosine": 0.9976232647895813 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.012148132819086313 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0691244974732399, + "mean_feedback_forward_cosine": 0.9994335142048922, + "mean_feedback_forward_relative_error": 0.03155574777417562, + "min_feedback_forward_cosine": 0.997624397277832 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.011397822160273791 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06910843402147293, + "mean_feedback_forward_cosine": 0.999433725530451, + "mean_feedback_forward_relative_error": 0.03155008294534954, + "min_feedback_forward_cosine": 0.9976255297660828 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.011458082374185323 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06909443438053131, + "mean_feedback_forward_cosine": 0.9994339130141519, + "mean_feedback_forward_relative_error": 0.0315452404066243, + "min_feedback_forward_cosine": 0.9976264834403992 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.009869020701497793 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06907776743173599, + "mean_feedback_forward_cosine": 0.9994341232559898, + "mean_feedback_forward_relative_error": 0.03153970285572789, + "min_feedback_forward_cosine": 0.9976276159286499 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.010547742835283279 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06906381249427795, + "mean_feedback_forward_cosine": 0.999434313990853, + "mean_feedback_forward_relative_error": 0.03153472960164601, + "min_feedback_forward_cosine": 0.9976285099983215 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.009300732849575579 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06904778629541397, + "mean_feedback_forward_cosine": 0.9994345090605996, + "mean_feedback_forward_relative_error": 0.031529135595668445, + "min_feedback_forward_cosine": 0.9976296424865723 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.009818964008167386 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06902909278869629, + "mean_feedback_forward_cosine": 0.999434745311737, + "mean_feedback_forward_relative_error": 0.03152294444258918, + "min_feedback_forward_cosine": 0.9976310133934021 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.010077409034781158 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06901176273822784, + "mean_feedback_forward_cosine": 0.9994349924000827, + "mean_feedback_forward_relative_error": 0.031517018733376805, + "min_feedback_forward_cosine": 0.9976321458816528 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.009160347877591849 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06899315118789673, + "mean_feedback_forward_cosine": 0.9994351983070373, + "mean_feedback_forward_relative_error": 0.031511178113181486, + "min_feedback_forward_cosine": 0.9976334571838379 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.008961532037444413 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06897760182619095, + "mean_feedback_forward_cosine": 0.9994353966279463, + "mean_feedback_forward_relative_error": 0.03150578120892698, + "min_feedback_forward_cosine": 0.9976345300674438 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.008542742440328002 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06896280497312546, + "mean_feedback_forward_cosine": 0.9994355873628096, + "mean_feedback_forward_relative_error": 0.031500711986287074, + "min_feedback_forward_cosine": 0.9976356029510498 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.008353168544620275 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06894773989915848, + "mean_feedback_forward_cosine": 0.9994357878511603, + "mean_feedback_forward_relative_error": 0.031495631257580085, + "min_feedback_forward_cosine": 0.9976365566253662 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.008146675640046596 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06893233209848404, + "mean_feedback_forward_cosine": 0.9994359666650946, + "mean_feedback_forward_relative_error": 0.03149048773233186, + "min_feedback_forward_cosine": 0.9976376295089722 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.007538513074964285 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06891530007123947, + "mean_feedback_forward_cosine": 0.9994361877441407, + "mean_feedback_forward_relative_error": 0.03148495936766267, + "min_feedback_forward_cosine": 0.9976388216018677 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.007756935535706579 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06889911741018295, + "mean_feedback_forward_cosine": 0.999436401237141, + "mean_feedback_forward_relative_error": 0.031479515169154516, + "min_feedback_forward_cosine": 0.9976398944854736 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.007756191560998559 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06888337433338165, + "mean_feedback_forward_cosine": 0.9994366049766541, + "mean_feedback_forward_relative_error": 0.031474347962913186, + "min_feedback_forward_cosine": 0.9976409077644348 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.0069661939964443444 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06886947154998779, + "mean_feedback_forward_cosine": 0.9994367707859386, + "mean_feedback_forward_relative_error": 0.03146959664300084, + "min_feedback_forward_cosine": 0.9976418614387512 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.006709710410907865 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0688507929444313, + "mean_feedback_forward_cosine": 0.9994369961998679, + "mean_feedback_forward_relative_error": 0.0314637197858908, + "min_feedback_forward_cosine": 0.9976431131362915 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.007294634993895889 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06883271783590317, + "mean_feedback_forward_cosine": 0.9994372302835638, + "mean_feedback_forward_relative_error": 0.03145776710219004, + "min_feedback_forward_cosine": 0.9976444244384766 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.007054433014690876 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06881632655858994, + "mean_feedback_forward_cosine": 0.9994374318556352, + "mean_feedback_forward_relative_error": 0.03145231891592795, + "min_feedback_forward_cosine": 0.9976456165313721 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.006983992516249418 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06879774481058121, + "mean_feedback_forward_cosine": 0.9994376886974682, + "mean_feedback_forward_relative_error": 0.03144586983729492, + "min_feedback_forward_cosine": 0.9976467490196228 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007579090554155409 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06878021359443665, + "mean_feedback_forward_cosine": 0.9994379151951183, + "mean_feedback_forward_relative_error": 0.0314398794604296, + "min_feedback_forward_cosine": 0.9976479411125183 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.007095893447734415 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06876447051763535, + "mean_feedback_forward_cosine": 0.9994381048462607, + "mean_feedback_forward_relative_error": 0.03143445560708642, + "min_feedback_forward_cosine": 0.9976490139961243 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006462747115045786 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06874611973762512, + "mean_feedback_forward_cosine": 0.999438354102048, + "mean_feedback_forward_relative_error": 0.03142826100811362, + "min_feedback_forward_cosine": 0.9976503252983093 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.006878289589509368 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06872718781232834, + "mean_feedback_forward_cosine": 0.9994386044415561, + "mean_feedback_forward_relative_error": 0.0314219257963652, + "min_feedback_forward_cosine": 0.9976516366004944 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.0068016401921957735 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06870897114276886, + "mean_feedback_forward_cosine": 0.9994388276880438, + "mean_feedback_forward_relative_error": 0.031415897768668155, + "min_feedback_forward_cosine": 0.9976528882980347 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.006183813361376524 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06869076192378998, + "mean_feedback_forward_cosine": 0.9994390563531356, + "mean_feedback_forward_relative_error": 0.031410239746963435, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005665142865814269 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06869052350521088, + "mean_feedback_forward_cosine": 0.999439059604298, + "mean_feedback_forward_relative_error": 0.0314101467163048, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.006256562349833548 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06869038194417953, + "mean_feedback_forward_cosine": 0.9994390639391812, + "mean_feedback_forward_relative_error": 0.03141007806936448, + "min_feedback_forward_cosine": 0.997654139995575 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.00507289287718013 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06869025528430939, + "mean_feedback_forward_cosine": 0.9994390574368563, + "mean_feedback_forward_relative_error": 0.03141001376577399, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0047390447936952115 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06869012117385864, + "mean_feedback_forward_cosine": 0.999439065022902, + "mean_feedback_forward_relative_error": 0.031409954017197546, + "min_feedback_forward_cosine": 0.9976540803909302 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004380627877153456 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868996471166611, + "mean_feedback_forward_cosine": 0.9994390661066229, + "mean_feedback_forward_relative_error": 0.031409888883883304, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004598957079797983 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868986040353775, + "mean_feedback_forward_cosine": 0.9994390498508107, + "mean_feedback_forward_relative_error": 0.031409834867173976, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.003838625859953463 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868971884250641, + "mean_feedback_forward_cosine": 0.9994390465996482, + "mean_feedback_forward_relative_error": 0.03140976984392513, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004432131025716662 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868959218263626, + "mean_feedback_forward_cosine": 0.9994390661066229, + "mean_feedback_forward_relative_error": 0.031409713769839566, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0044609378554672004 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868948042392731, + "mean_feedback_forward_cosine": 0.9994390715252269, + "mean_feedback_forward_relative_error": 0.03140965415672822, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.004222902919054031 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868936866521835, + "mean_feedback_forward_cosine": 0.9994390617717396, + "mean_feedback_forward_relative_error": 0.031409599987620655, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.003845143410935998 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868921965360641, + "mean_feedback_forward_cosine": 0.9994390563531356, + "mean_feedback_forward_relative_error": 0.031409540425308725, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004378214224409312 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868909299373627, + "mean_feedback_forward_cosine": 0.9994390617717396, + "mean_feedback_forward_relative_error": 0.031409482336179774, + "min_feedback_forward_cosine": 0.9976541996002197 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.004232344762831926 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868896633386612, + "mean_feedback_forward_cosine": 0.9994390661066229, + "mean_feedback_forward_relative_error": 0.03140942423858426, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.004272423257436603 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868885457515717, + "mean_feedback_forward_cosine": 0.9994390736926686, + "mean_feedback_forward_relative_error": 0.03140936523506587, + "min_feedback_forward_cosine": 0.9976542592048645 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.004210389293860644 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868872046470642, + "mean_feedback_forward_cosine": 0.999439076943831, + "mean_feedback_forward_relative_error": 0.03140930999070406, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.003977990410029888 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868861615657806, + "mean_feedback_forward_cosine": 0.9994390693577853, + "mean_feedback_forward_relative_error": 0.031409247414293615, + "min_feedback_forward_cosine": 0.9976542592048645 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.004078018093742431 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868851184844971, + "mean_feedback_forward_cosine": 0.9994390921159224, + "mean_feedback_forward_relative_error": 0.031409196327017115, + "min_feedback_forward_cosine": 0.9976542592048645 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003648907724171877 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868840754032135, + "mean_feedback_forward_cosine": 0.9994390834461558, + "mean_feedback_forward_relative_error": 0.03140914492647756, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.0033877091257832944 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0686882957816124, + "mean_feedback_forward_cosine": 0.9994390823624351, + "mean_feedback_forward_relative_error": 0.03140909403393215, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0035693821069598198 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868819147348404, + "mean_feedback_forward_cosine": 0.9994390866973183, + "mean_feedback_forward_relative_error": 0.03140904228626327, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0036892011823877694 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868808716535568, + "mean_feedback_forward_cosine": 0.9994390791112726, + "mean_feedback_forward_relative_error": 0.03140899477187883, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.003492035176092759 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868799030780792, + "mean_feedback_forward_cosine": 0.9994390877810392, + "mean_feedback_forward_relative_error": 0.03140894459052519, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003594661014340818 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868788599967957, + "mean_feedback_forward_cosine": 0.9994390856135975, + "mean_feedback_forward_relative_error": 0.031408889811824665, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.003707864111084491 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868778169155121, + "mean_feedback_forward_cosine": 0.99943908886476, + "mean_feedback_forward_relative_error": 0.03140884155238217, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.003334373738132417 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868770718574524, + "mean_feedback_forward_cosine": 0.9994390986182473, + "mean_feedback_forward_relative_error": 0.031408793199807404, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.003135205378085375 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868759542703629, + "mean_feedback_forward_cosine": 0.9994390921159224, + "mean_feedback_forward_relative_error": 0.031408739513294266, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.00379005764760077 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868747621774673, + "mean_feedback_forward_cosine": 0.9994390856135975, + "mean_feedback_forward_relative_error": 0.03140868591991338, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0037669762847572564 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868737190961838, + "mean_feedback_forward_cosine": 0.9994390921159224, + "mean_feedback_forward_relative_error": 0.03140863081948324, + "min_feedback_forward_cosine": 0.9976543188095093 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0033926852117478847 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868726015090942, + "mean_feedback_forward_cosine": 0.9994390975345265, + "mean_feedback_forward_relative_error": 0.03140857703137127, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.003614906167127192 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868716329336166, + "mean_feedback_forward_cosine": 0.9994390931996432, + "mean_feedback_forward_relative_error": 0.03140852423384786, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0034720032831281424 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868704408407211, + "mean_feedback_forward_cosine": 0.9994391029531305, + "mean_feedback_forward_relative_error": 0.031408465721390465, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0035889510291069746 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868693232536316, + "mean_feedback_forward_cosine": 0.9994390997019681, + "mean_feedback_forward_relative_error": 0.03140841539610516, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.003305433737561107 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0686868354678154, + "mean_feedback_forward_cosine": 0.9994391202926636, + "mean_feedback_forward_relative_error": 0.03140837007456205, + "min_feedback_forward_cosine": 0.997654378414154 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.003175204580221325 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868674606084824, + "mean_feedback_forward_cosine": 0.9994391170415011, + "mean_feedback_forward_relative_error": 0.03140832256017761, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0031294494361244142 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868667155504227, + "mean_feedback_forward_cosine": 0.9994391116228971, + "mean_feedback_forward_relative_error": 0.03140827675604008, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.003196823445223272 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0686865746974945, + "mean_feedback_forward_cosine": 0.9994391029531305, + "mean_feedback_forward_relative_error": 0.03140823346647349, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0030139125097729267 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868649274110794, + "mean_feedback_forward_cosine": 0.9994390910322016, + "mean_feedback_forward_relative_error": 0.03140818929638375, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.003099756380841136 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868639588356018, + "mean_feedback_forward_cosine": 0.9994391040368513, + "mean_feedback_forward_relative_error": 0.031408142391592266, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0031279313344694673 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868629902601242, + "mean_feedback_forward_cosine": 0.9994391029531305, + "mean_feedback_forward_relative_error": 0.03140809452161193, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.0032867785268090666 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868621706962585, + "mean_feedback_forward_cosine": 0.9994391105391762, + "mean_feedback_forward_relative_error": 0.03140805058858611, + "min_feedback_forward_cosine": 0.9976544976234436 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0030857411340996623 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0686860904097557, + "mean_feedback_forward_cosine": 0.9994391148740595, + "mean_feedback_forward_relative_error": 0.03140799194066362, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.004001324234195054 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868598610162735, + "mean_feedback_forward_cosine": 0.9994391094554554, + "mean_feedback_forward_relative_error": 0.031407940692522306, + "min_feedback_forward_cosine": 0.9976544380187988 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.003275472015310079 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868590414524078, + "mean_feedback_forward_cosine": 0.9994391170415011, + "mean_feedback_forward_relative_error": 0.03140789498151703, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.002768472644649446 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868579983711243, + "mean_feedback_forward_cosine": 0.9994391202926636, + "mean_feedback_forward_relative_error": 0.03140784820372408, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0030618526208773256 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868570297956467, + "mean_feedback_forward_cosine": 0.9994391040368513, + "mean_feedback_forward_relative_error": 0.03140779734504494, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.00339243379432708 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0686856135725975, + "mean_feedback_forward_cosine": 0.999439123543826, + "mean_feedback_forward_relative_error": 0.031407752802426164, + "min_feedback_forward_cosine": 0.9976546168327332 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0028797079719975593 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868550926446915, + "mean_feedback_forward_cosine": 0.9994391224601052, + "mean_feedback_forward_relative_error": 0.03140770508484407, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0033768319156579673 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868541985750198, + "mean_feedback_forward_cosine": 0.9994391332973134, + "mean_feedback_forward_relative_error": 0.03140765925530683, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0029052105254959316 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868532299995422, + "mean_feedback_forward_cosine": 0.9994391322135925, + "mean_feedback_forward_relative_error": 0.03140761311250654, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.003043961836770177 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06868523359298706, + "mean_feedback_forward_cosine": 0.9994391322135925, + "mean_feedback_forward_relative_error": 0.031407567316835576, + "min_feedback_forward_cosine": 0.9976545572280884 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0028966831861622633 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9265, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3951418914794922, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3318545920, + "peak_memory_reserved_bytes": 4754243584, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0679945945739746, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 18253.185290575027, + "train_wall_s": 18247.471817970276 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250433280000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 5010580480000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json new file mode 100644 index 0000000..2afc87e --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json @@ -0,0 +1,3654 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9995148744848039, + "feedback_forward_cosine": [ + 0.9997456073760986, + 0.9997330904006958, + 0.9995865821838379, + 0.9995840787887573, + 0.9996457099914551, + 0.9996064901351929, + 0.9997146129608154, + 0.9996328353881836, + 0.999509334564209, + 0.9994516968727112, + 0.9995710849761963, + 0.9995086193084717, + 0.9996844530105591, + 0.9996521472930908, + 0.9997467994689941, + 0.9996289014816284, + 0.9997332692146301, + 0.9996172785758972, + 0.9996885657310486, + 0.9997808337211609, + 0.9997159242630005, + 0.9996783137321472, + 0.999729335308075, + 0.999666154384613, + 0.9997511506080627, + 0.999657154083252, + 0.9996792078018188, + 0.9995862245559692, + 0.9996955394744873, + 0.9995726346969604, + 0.9995874166488647, + 0.9994576573371887, + 0.9996607303619385, + 0.9995304346084595, + 0.9995937943458557, + 0.9994197487831116, + 0.9997514486312866, + 0.9998229742050171, + 0.9997981190681458, + 0.9997464418411255, + 0.9997518062591553, + 0.9996287226676941, + 0.9996957778930664, + 0.9995288848876953, + 0.9996234774589539, + 0.999355673789978, + 0.9994724988937378, + 0.9991140961647034, + 0.9988570213317871, + 0.9983136057853699, + 0.9983664751052856, + 0.9977402687072754, + 0.9987434148788452, + 0.9979894161224365, + 0.9999787211418152 + ], + "feedback_forward_norm_ratio": [ + 0.9988730549812317, + 1.0006078481674194, + 0.9993394613265991, + 1.001127004623413, + 0.9990338683128357, + 1.0004154443740845, + 0.9995367527008057, + 1.0011768341064453, + 0.9985411167144775, + 1.0020065307617188, + 0.9981188178062439, + 1.0027170181274414, + 0.9990379214286804, + 1.0009950399398804, + 0.9998155832290649, + 1.0005804300308228, + 0.9991078972816467, + 1.0011104345321655, + 0.9990886449813843, + 1.0007953643798828, + 0.9995934963226318, + 1.000977635383606, + 0.9990978837013245, + 1.0008434057235718, + 0.9994692802429199, + 1.0015718936920166, + 0.9989431500434875, + 1.0015720129013062, + 0.9993394017219543, + 1.001583456993103, + 0.9989436864852905, + 1.0021345615386963, + 0.9989631175994873, + 1.001844048500061, + 0.9989639520645142, + 1.003008484840393, + 0.9993592500686646, + 1.0002747774124146, + 0.9998732805252075, + 1.00054931640625, + 0.9995817542076111, + 1.0010743141174316, + 0.999226450920105, + 1.0015519857406616, + 0.9992843270301819, + 1.0018341541290283, + 0.998745322227478, + 1.0020575523376465, + 0.9983550906181335, + 1.0026588439941406, + 0.9973151087760925, + 1.0030194520950317, + 0.9985994100570679, + 1.0021740198135376, + 1.001080870628357 + ], + "feedback_forward_relative_error": [ + 0.022569570690393448, + 0.023116396740078926, + 0.028753137215971947, + 0.02887820266187191, + 0.02662196196615696, + 0.028064891695976257, + 0.023886168375611305, + 0.027140865102410316, + 0.03134024888277054, + 0.033206596970558167, + 0.02931816317141056, + 0.031509820371866226, + 0.025130072608590126, + 0.026406964287161827, + 0.02250189147889614, + 0.02725551277399063, + 0.02310079149901867, + 0.027704205363988876, + 0.024963723495602608, + 0.02095712162554264, + 0.023833798244595528, + 0.025395900011062622, + 0.02327672205865383, + 0.0258642490953207, + 0.022308772429823875, + 0.02625497803092003, + 0.025336265563964844, + 0.028834355995059013, + 0.024675320833921432, + 0.0293009914457798, + 0.028729842975735664, + 0.033037349581718445, + 0.02605678141117096, + 0.030729036778211594, + 0.02850693091750145, + 0.03424990922212601, + 0.022296937182545662, + 0.01881887950003147, + 0.020092979073524475, + 0.02253291755914688, + 0.022278066724538803, + 0.027286378666758537, + 0.024672068655490875, + 0.03076157532632351, + 0.027442889288067818, + 0.035978950560092926, + 0.03248658403754234, + 0.04218292236328125, + 0.04779919609427452, + 0.058212172240018845, + 0.057143598794937134, + 0.06739622354507446, + 0.05011552944779396, + 0.06351963430643082, + 0.006617867853492498 + ], + "innovation_negative_gradient_cosine": [ + 0.9995577931404114, + 0.9995124340057373, + 0.9995529055595398, + 0.9994076490402222, + 0.999546468257904, + 0.9993767738342285, + 0.9995553493499756, + 0.9994248151779175, + 0.9995616674423218, + 0.9994311332702637, + 0.9995665550231934, + 0.9993959665298462, + 0.9995825886726379, + 0.9995402097702026, + 0.9995884895324707, + 0.9995632767677307, + 0.9995993375778198, + 0.9995043277740479, + 0.999602735042572, + 0.999591588973999, + 0.9996074438095093, + 0.9995112419128418, + 0.9996109008789062, + 0.9995533227920532, + 0.9996251463890076, + 0.9995191693305969, + 0.9996445178985596, + 0.9995243549346924, + 0.9996548891067505, + 0.9995096921920776, + 0.999665379524231, + 0.9994922876358032, + 0.9996774792671204, + 0.9995546340942383, + 0.9996901154518127, + 0.9994900226593018, + 0.9996975660324097, + 0.9997286200523376, + 0.9997000694274902, + 0.9996534585952759, + 0.9996994137763977, + 0.9996539354324341, + 0.9997164011001587, + 0.9996889233589172, + 0.9997352957725525, + 0.9996839761734009, + 0.9997507333755493, + 0.9997549057006836, + 0.9996919631958008, + 0.9995550513267517, + 0.9996441602706909, + 0.999585747718811, + 0.9998013973236084, + 0.9998618364334106, + 0.7703130841255188 + ], + "mean_feedback_forward_cosine": 0.9994924046776511, + "mean_feedback_forward_relative_error": 0.0299355069598691, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9995577931404114, + 0.9995124340057373, + 0.9995529055595398, + 0.9994076490402222, + 0.999546468257904, + 0.9993767738342285, + 0.9995553493499756, + 0.9994248151779175, + 0.9995616674423218, + 0.9994311332702637, + 0.9995665550231934, + 0.9993959665298462, + 0.9995825886726379, + 0.9995402097702026, + 0.9995884895324707, + 0.9995632767677307, + 0.9995993375778198, + 0.9995043277740479, + 0.999602735042572, + 0.999591588973999, + 0.9996074438095093, + 0.9995112419128418, + 0.9996109008789062, + 0.9995533227920532, + 0.9996251463890076, + 0.9995191693305969, + 0.9996445178985596, + 0.9995243549346924, + 0.9996548891067505, + 0.9995096921920776, + 0.999665379524231, + 0.9994922876358032, + 0.9996774792671204, + 0.9995546340942383, + 0.9996901154518127, + 0.9994900226593018, + 0.9996975660324097, + 0.9997286200523376, + 0.9997000694274902, + 0.9996534585952759, + 0.9996994137763977, + 0.9996539354324341, + 0.9997164011001587, + 0.9996889233589172, + 0.9997352957725525, + 0.9996839761734009, + 0.9997507333755493, + 0.9997549057006836, + 0.9996919631958008, + 0.9995550513267517, + 0.9996441602706909, + 0.999585747718811, + 0.9998013973236084, + 0.9998618364334106, + 0.7703130841255188 + ], + "teaching_negative_gradient_cosine": [ + 0.9995577931404114, + 0.9995124340057373, + 0.9995529055595398, + 0.9994076490402222, + 0.999546468257904, + 0.9993767738342285, + 0.9995553493499756, + 0.9994248151779175, + 0.9995616674423218, + 0.9994311332702637, + 0.9995665550231934, + 0.9993959665298462, + 0.9995825886726379, + 0.9995402097702026, + 0.9995884895324707, + 0.9995632767677307, + 0.9995993375778198, + 0.9995043277740479, + 0.999602735042572, + 0.999591588973999, + 0.9996074438095093, + 0.9995112419128418, + 0.9996109008789062, + 0.9995533227920532, + 0.9996251463890076, + 0.9995191693305969, + 0.9996445178985596, + 0.9995243549346924, + 0.9996548891067505, + 0.9995096921920776, + 0.999665379524231, + 0.9994922876358032, + 0.9996774792671204, + 0.9995546340942383, + 0.9996901154518127, + 0.9994900226593018, + 0.9996975660324097, + 0.9997286200523376, + 0.9997000694274902, + 0.9996534585952759, + 0.9996994137763977, + 0.9996539354324341, + 0.9997164011001587, + 0.9996889233589172, + 0.9997352957725525, + 0.9996839761734009, + 0.9997507333755493, + 0.9997549057006836, + 0.9996919631958008, + 0.9995550513267517, + 0.9996441602706909, + 0.999585747718811, + 0.9998013973236084, + 0.9998618364334106, + 0.7703130841255188 + ], + "wall_s": 0.1903843879699707 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3721873760223389, + "mean_feedback_forward_cosine": 0.22906434779817408, + "mean_feedback_forward_relative_error": 1.23561264926737, + "min_feedback_forward_cosine": 0.06822853535413742 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.618657088623047 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3618451356887817, + "mean_feedback_forward_cosine": 0.2514899350025437, + "mean_feedback_forward_relative_error": 1.216604634848508, + "min_feedback_forward_cosine": 0.0825369656085968 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 2.0443328186035155 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3392257690429688, + "mean_feedback_forward_cosine": 0.27437169619581914, + "mean_feedback_forward_relative_error": 1.1974607321349058, + "min_feedback_forward_cosine": 0.1043594554066658 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.8029051454925538 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3336762189865112, + "mean_feedback_forward_cosine": 0.3102063487876545, + "mean_feedback_forward_relative_error": 1.1679667462002148, + "min_feedback_forward_cosine": 0.11178690195083618 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.5733120203781128 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3269165754318237, + "mean_feedback_forward_cosine": 0.3568539035591212, + "mean_feedback_forward_relative_error": 1.1278487942435524, + "min_feedback_forward_cosine": 0.11940619349479675 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.3117623218536376 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3227629661560059, + "mean_feedback_forward_cosine": 0.4043669891628352, + "mean_feedback_forward_relative_error": 1.085144471580332, + "min_feedback_forward_cosine": 0.12460734695196152 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 1.0928271724700929 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3171675205230713, + "mean_feedback_forward_cosine": 0.4497559509494088, + "mean_feedback_forward_relative_error": 1.0418285944245078, + "min_feedback_forward_cosine": 0.13213735818862915 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.9260222872543334 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3109973669052124, + "mean_feedback_forward_cosine": 0.4906862039457668, + "mean_feedback_forward_relative_error": 1.0011875304308804, + "min_feedback_forward_cosine": 0.14014112949371338 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.8086126829910278 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3044464588165283, + "mean_feedback_forward_cosine": 0.5249264020811427, + "mean_feedback_forward_relative_error": 0.9653921685435555, + "min_feedback_forward_cosine": 0.14885416626930237 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.7224795008468627 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2969162464141846, + "mean_feedback_forward_cosine": 0.555767253854058, + "mean_feedback_forward_relative_error": 0.9317991077899933, + "min_feedback_forward_cosine": 0.1582750529050827 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.6612757705783844 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2884743213653564, + "mean_feedback_forward_cosine": 0.5836254799907857, + "mean_feedback_forward_relative_error": 0.9001913368701935, + "min_feedback_forward_cosine": 0.1689441055059433 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.6117791198348999 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2787386178970337, + "mean_feedback_forward_cosine": 0.609176770936359, + "mean_feedback_forward_relative_error": 0.8700118612159382, + "min_feedback_forward_cosine": 0.18143343925476074 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.574167943572998 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.268519401550293, + "mean_feedback_forward_cosine": 0.6319928404959765, + "mean_feedback_forward_relative_error": 0.8422089395197955, + "min_feedback_forward_cosine": 0.19407938420772552 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.5307482202529907 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2560867071151733, + "mean_feedback_forward_cosine": 0.6534666595133868, + "mean_feedback_forward_relative_error": 0.8151899307966233, + "min_feedback_forward_cosine": 0.20934352278709412 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.5125123575782776 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2439846992492676, + "mean_feedback_forward_cosine": 0.6732584809715098, + "mean_feedback_forward_relative_error": 0.7894332414323634, + "min_feedback_forward_cosine": 0.22344081103801727 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4776908352661133 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2297369241714478, + "mean_feedback_forward_cosine": 0.6926673897288063, + "mean_feedback_forward_relative_error": 0.7637290185148066, + "min_feedback_forward_cosine": 0.2404581755399704 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.46589656856536865 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2135229110717773, + "mean_feedback_forward_cosine": 0.7107760201800953, + "mean_feedback_forward_relative_error": 0.7388164438984611, + "min_feedback_forward_cosine": 0.2596164643764496 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.44467153213500976 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.197867512702942, + "mean_feedback_forward_cosine": 0.7274938036095012, + "mean_feedback_forward_relative_error": 0.7151795102791353, + "min_feedback_forward_cosine": 0.27816063165664673 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.42247454542160034 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1803089380264282, + "mean_feedback_forward_cosine": 0.7433888825503263, + "mean_feedback_forward_relative_error": 0.6923050625757737, + "min_feedback_forward_cosine": 0.2987276315689087 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.40903840388298035 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.164432168006897, + "mean_feedback_forward_cosine": 0.7582452020861886, + "mean_feedback_forward_relative_error": 0.6701002207669344, + "min_feedback_forward_cosine": 0.31991180777549744 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3893443667793274 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1477042436599731, + "mean_feedback_forward_cosine": 0.7729029227386821, + "mean_feedback_forward_relative_error": 0.6478099530393427, + "min_feedback_forward_cosine": 0.34410470724105835 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.37688736102104187 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1307538747787476, + "mean_feedback_forward_cosine": 0.7863774174993688, + "mean_feedback_forward_relative_error": 0.6267310367389158, + "min_feedback_forward_cosine": 0.36850911378860474 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.36979484375 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1123244762420654, + "mean_feedback_forward_cosine": 0.7996666431427002, + "mean_feedback_forward_relative_error": 0.6054776159199802, + "min_feedback_forward_cosine": 0.39363932609558105 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3512759153175354 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0937360525131226, + "mean_feedback_forward_cosine": 0.8120680473067544, + "mean_feedback_forward_relative_error": 0.584982961958105, + "min_feedback_forward_cosine": 0.41800665855407715 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3465504659366608 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0731472969055176, + "mean_feedback_forward_cosine": 0.823938865553249, + "mean_feedback_forward_relative_error": 0.5647742089900103, + "min_feedback_forward_cosine": 0.44163089990615845 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3362824455356598 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.054181694984436, + "mean_feedback_forward_cosine": 0.8349253486503254, + "mean_feedback_forward_relative_error": 0.5455743025649678, + "min_feedback_forward_cosine": 0.46197107434272766 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3249240168380737 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0323097705841064, + "mean_feedback_forward_cosine": 0.8456882639364762, + "mean_feedback_forward_relative_error": 0.5262904961000789, + "min_feedback_forward_cosine": 0.4847239851951599 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.31498544347763063 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0116260051727295, + "mean_feedback_forward_cosine": 0.8558149587024342, + "mean_feedback_forward_relative_error": 0.507576184110208, + "min_feedback_forward_cosine": 0.5057708024978638 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3097911694049835 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9895654916763306, + "mean_feedback_forward_cosine": 0.8650211746042425, + "mean_feedback_forward_relative_error": 0.49017953926866703, + "min_feedback_forward_cosine": 0.5277377367019653 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.29969036979675295 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9655799269676208, + "mean_feedback_forward_cosine": 0.8739804907278581, + "mean_feedback_forward_relative_error": 0.47264781160788105, + "min_feedback_forward_cosine": 0.5506198406219482 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.29819255435943604 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9415902495384216, + "mean_feedback_forward_cosine": 0.8824259974739769, + "mean_feedback_forward_relative_error": 0.45579604953527453, + "min_feedback_forward_cosine": 0.5728276371955872 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2815729883956909 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9179940223693848, + "mean_feedback_forward_cosine": 0.8905050201849504, + "mean_feedback_forward_relative_error": 0.43920987912199716, + "min_feedback_forward_cosine": 0.5942811369895935 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.28759058656692504 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.894770085811615, + "mean_feedback_forward_cosine": 0.8977142171426253, + "mean_feedback_forward_relative_error": 0.423809283294461, + "min_feedback_forward_cosine": 0.6146277785301208 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2714368555927277 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8704290390014648, + "mean_feedback_forward_cosine": 0.9047920725562356, + "mean_feedback_forward_relative_error": 0.40825533162463795, + "min_feedback_forward_cosine": 0.6355072259902954 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2704858200263977 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8463582992553711, + "mean_feedback_forward_cosine": 0.9114495082335039, + "mean_feedback_forward_relative_error": 0.393126321922649, + "min_feedback_forward_cosine": 0.6554253101348877 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2692690270805359 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8230829834938049, + "mean_feedback_forward_cosine": 0.9175222591920332, + "mean_feedback_forward_relative_error": 0.3787562095306136, + "min_feedback_forward_cosine": 0.67411869764328 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25900745127677915 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7997261881828308, + "mean_feedback_forward_cosine": 0.9232426578348333, + "mean_feedback_forward_relative_error": 0.36491667099974373, + "min_feedback_forward_cosine": 0.6921552419662476 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2505539571809769 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7771807909011841, + "mean_feedback_forward_cosine": 0.928544446555051, + "mean_feedback_forward_relative_error": 0.35170431380922146, + "min_feedback_forward_cosine": 0.709445059299469 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2463171188735962 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7533586621284485, + "mean_feedback_forward_cosine": 0.933710492740978, + "mean_feedback_forward_relative_error": 0.3383773974396966, + "min_feedback_forward_cosine": 0.726951539516449 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2437849562740326 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7309722304344177, + "mean_feedback_forward_cosine": 0.9382276155731895, + "mean_feedback_forward_relative_error": 0.3263433360240676, + "min_feedback_forward_cosine": 0.7429503798484802 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23817058839797975 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7088060975074768, + "mean_feedback_forward_cosine": 0.9425449761477384, + "mean_feedback_forward_relative_error": 0.31445055292411284, + "min_feedback_forward_cosine": 0.7582069635391235 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.23429553953170776 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6824594140052795, + "mean_feedback_forward_cosine": 0.9470177336172624, + "mean_feedback_forward_relative_error": 0.3017793464389714, + "min_feedback_forward_cosine": 0.7757521867752075 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23840826580524443 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6601223349571228, + "mean_feedback_forward_cosine": 0.9508190794424577, + "mean_feedback_forward_relative_error": 0.2904875550757755, + "min_feedback_forward_cosine": 0.7900207042694092 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22959980630874635 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6384415030479431, + "mean_feedback_forward_cosine": 0.9544151338663969, + "mean_feedback_forward_relative_error": 0.27948091301051053, + "min_feedback_forward_cosine": 0.8036364316940308 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.23084515382766724 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6156907677650452, + "mean_feedback_forward_cosine": 0.9577961065552452, + "mean_feedback_forward_relative_error": 0.26881950958208606, + "min_feedback_forward_cosine": 0.8170660138130188 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22511683757781983 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5947313904762268, + "mean_feedback_forward_cosine": 0.9608323324810375, + "mean_feedback_forward_relative_error": 0.25886820771477437, + "min_feedback_forward_cosine": 0.8291815519332886 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21716432525157928 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5736299753189087, + "mean_feedback_forward_cosine": 0.9637927716428584, + "mean_feedback_forward_relative_error": 0.24875448203899644, + "min_feedback_forward_cosine": 0.8410825133323669 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2182045086479187 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5530214905738831, + "mean_feedback_forward_cosine": 0.9664847081357782, + "mean_feedback_forward_relative_error": 0.23927537900480358, + "min_feedback_forward_cosine": 0.8522039651870728 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21188023831367492 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5337027907371521, + "mean_feedback_forward_cosine": 0.968951736796986, + "mean_feedback_forward_relative_error": 0.2302019331942905, + "min_feedback_forward_cosine": 0.8622238636016846 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20758117275238036 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5153356790542603, + "mean_feedback_forward_cosine": 0.9713034532286904, + "mean_feedback_forward_relative_error": 0.22121331122788515, + "min_feedback_forward_cosine": 0.8714470863342285 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2128644209384918 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4960263669490814, + "mean_feedback_forward_cosine": 0.9734993869608098, + "mean_feedback_forward_relative_error": 0.21256887032227081, + "min_feedback_forward_cosine": 0.8808215260505676 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20747516375541686 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4773900806903839, + "mean_feedback_forward_cosine": 0.9755356550216675, + "mean_feedback_forward_relative_error": 0.2042575067417188, + "min_feedback_forward_cosine": 0.8895806074142456 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20645665825366974 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4604026973247528, + "mean_feedback_forward_cosine": 0.9773342045870694, + "mean_feedback_forward_relative_error": 0.19654388956048272, + "min_feedback_forward_cosine": 0.8970670700073242 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20073466979026794 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4430951476097107, + "mean_feedback_forward_cosine": 0.9790165261788801, + "mean_feedback_forward_relative_error": 0.1891219674186273, + "min_feedback_forward_cosine": 0.904595136642456 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19337763548851014 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42659634351730347, + "mean_feedback_forward_cosine": 0.9805894006382335, + "mean_feedback_forward_relative_error": 0.18193596174771134, + "min_feedback_forward_cosine": 0.9115451574325562 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.193649558801651 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4102657735347748, + "mean_feedback_forward_cosine": 0.9820752284743569, + "mean_feedback_forward_relative_error": 0.1748588041825728, + "min_feedback_forward_cosine": 0.91810142993927 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1931214778614044 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3945562541484833, + "mean_feedback_forward_cosine": 0.9834433187137951, + "mean_feedback_forward_relative_error": 0.16806023865938186, + "min_feedback_forward_cosine": 0.9241763353347778 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19203194622039796 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.380015105009079, + "mean_feedback_forward_cosine": 0.9846876285292886, + "mean_feedback_forward_relative_error": 0.16161761378700082, + "min_feedback_forward_cosine": 0.9296014904975891 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.18920780365467071 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36605533957481384, + "mean_feedback_forward_cosine": 0.9858385129408402, + "mean_feedback_forward_relative_error": 0.1554398418150165, + "min_feedback_forward_cosine": 0.9346195459365845 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.1875186634349823 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35228097438812256, + "mean_feedback_forward_cosine": 0.9869261893359098, + "mean_feedback_forward_relative_error": 0.1493348742750558, + "min_feedback_forward_cosine": 0.939432680606842 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18835279341697692 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33810895681381226, + "mean_feedback_forward_cosine": 0.9879302880980751, + "mean_feedback_forward_relative_error": 0.14352331920103592, + "min_feedback_forward_cosine": 0.9441512823104858 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18333342970848082 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3247024714946747, + "mean_feedback_forward_cosine": 0.9888717066157948, + "mean_feedback_forward_relative_error": 0.1378449701450088, + "min_feedback_forward_cosine": 0.9484421014785767 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.1856263377380371 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31059858202934265, + "mean_feedback_forward_cosine": 0.9897600022229281, + "mean_feedback_forward_relative_error": 0.13233231245116753, + "min_feedback_forward_cosine": 0.9527953267097473 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18473754248142243 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2986745834350586, + "mean_feedback_forward_cosine": 0.9905232527039268, + "mean_feedback_forward_relative_error": 0.1273461725224148, + "min_feedback_forward_cosine": 0.9563075304031372 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17438865018367766 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2868397831916809, + "mean_feedback_forward_cosine": 0.9912605014714327, + "mean_feedback_forward_relative_error": 0.12232954075390642, + "min_feedback_forward_cosine": 0.9596667289733887 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.180476394944191 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27480974793434143, + "mean_feedback_forward_cosine": 0.9919500871138139, + "mean_feedback_forward_relative_error": 0.11744141053747047, + "min_feedback_forward_cosine": 0.9629510641098022 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18252969101667404 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26383450627326965, + "mean_feedback_forward_cosine": 0.9925798947160894, + "mean_feedback_forward_relative_error": 0.11280005262656645, + "min_feedback_forward_cosine": 0.9658384323120117 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.17417837971687317 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25343167781829834, + "mean_feedback_forward_cosine": 0.9931712020527232, + "mean_feedback_forward_relative_error": 0.10825635316697034, + "min_feedback_forward_cosine": 0.9684686064720154 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18076635927200319 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24370114505290985, + "mean_feedback_forward_cosine": 0.9936928879130971, + "mean_feedback_forward_relative_error": 0.10406356636773456, + "min_feedback_forward_cosine": 0.9708285927772522 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.17136417557716369 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2343776673078537, + "mean_feedback_forward_cosine": 0.9941819310188293, + "mean_feedback_forward_relative_error": 0.09996809884905815, + "min_feedback_forward_cosine": 0.9730112552642822 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17325837555885315 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22468312084674835, + "mean_feedback_forward_cosine": 0.9946417169137435, + "mean_feedback_forward_relative_error": 0.09599588062952866, + "min_feedback_forward_cosine": 0.9751811027526855 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17303499470472336 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2157231569290161, + "mean_feedback_forward_cosine": 0.9950508767908269, + "mean_feedback_forward_relative_error": 0.09230448867787014, + "min_feedback_forward_cosine": 0.977104663848877 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16835947505950927 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20719926059246063, + "mean_feedback_forward_cosine": 0.9954314134337685, + "mean_feedback_forward_relative_error": 0.0887195079841397, + "min_feedback_forward_cosine": 0.978866696357727 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16926348195552826 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19898490607738495, + "mean_feedback_forward_cosine": 0.995786114172502, + "mean_feedback_forward_relative_error": 0.08523530509661545, + "min_feedback_forward_cosine": 0.9804971218109131 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16732301572322844 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1906488537788391, + "mean_feedback_forward_cosine": 0.9961130423979325, + "mean_feedback_forward_relative_error": 0.0818956772712144, + "min_feedback_forward_cosine": 0.9820825457572937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1667278288602829 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18334421515464783, + "mean_feedback_forward_cosine": 0.9964060230688615, + "mean_feedback_forward_relative_error": 0.07877858199856498, + "min_feedback_forward_cosine": 0.9834163188934326 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.16291458018779756 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17592990398406982, + "mean_feedback_forward_cosine": 0.9966853965412487, + "mean_feedback_forward_relative_error": 0.07569103329019113, + "min_feedback_forward_cosine": 0.9847314357757568 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.1661173464488983 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16859185695648193, + "mean_feedback_forward_cosine": 0.9969464638016441, + "mean_feedback_forward_relative_error": 0.07268435887315057, + "min_feedback_forward_cosine": 0.9859713315963745 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1632753385400772 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16158096492290497, + "mean_feedback_forward_cosine": 0.9971798214045438, + "mean_feedback_forward_relative_error": 0.06990636316212741, + "min_feedback_forward_cosine": 0.9871110916137695 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15748083257198334 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15530893206596375, + "mean_feedback_forward_cosine": 0.9974013209342957, + "mean_feedback_forward_relative_error": 0.06712234782224352, + "min_feedback_forward_cosine": 0.9880821704864502 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1633152895450592 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14913444221019745, + "mean_feedback_forward_cosine": 0.9976030414754694, + "mean_feedback_forward_relative_error": 0.06448946229436181, + "min_feedback_forward_cosine": 0.9890108108520508 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16058050225019455 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14306572079658508, + "mean_feedback_forward_cosine": 0.9977879004044966, + "mean_feedback_forward_relative_error": 0.06198189661584117, + "min_feedback_forward_cosine": 0.9898815155029297 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1555505397939682 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13722650706768036, + "mean_feedback_forward_cosine": 0.9979615753347223, + "mean_feedback_forward_relative_error": 0.05952306567606601, + "min_feedback_forward_cosine": 0.990685760974884 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15881990154266357 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13209199905395508, + "mean_feedback_forward_cosine": 0.9981146498159928, + "mean_feedback_forward_relative_error": 0.0572621196677739, + "min_feedback_forward_cosine": 0.9913676381111145 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1541928218650818 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1266280859708786, + "mean_feedback_forward_cosine": 0.9982635487209667, + "mean_feedback_forward_relative_error": 0.054980828579176556, + "min_feedback_forward_cosine": 0.9920614957809448 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15678585725784303 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12115006148815155, + "mean_feedback_forward_cosine": 0.9984057946638627, + "mean_feedback_forward_relative_error": 0.05271399278532375, + "min_feedback_forward_cosine": 0.9927307367324829 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1581881642627716 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11628136783838272, + "mean_feedback_forward_cosine": 0.9985274737531489, + "mean_feedback_forward_relative_error": 0.05068704489279877, + "min_feedback_forward_cosine": 0.9933011531829834 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15509069224357605 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1117628961801529, + "mean_feedback_forward_cosine": 0.9986399585550482, + "mean_feedback_forward_relative_error": 0.04872960036789829, + "min_feedback_forward_cosine": 0.9938088655471802 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.15393281336784362 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10720320045948029, + "mean_feedback_forward_cosine": 0.998745450106534, + "mean_feedback_forward_relative_error": 0.04682655361565677, + "min_feedback_forward_cosine": 0.9943026900291443 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1532562497806549 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10307206958532333, + "mean_feedback_forward_cosine": 0.9988405346870423, + "mean_feedback_forward_relative_error": 0.04503027690066533, + "min_feedback_forward_cosine": 0.9947317838668823 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.15164671294689178 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0990542471408844, + "mean_feedback_forward_cosine": 0.9989296945658597, + "mean_feedback_forward_relative_error": 0.043277259001677684, + "min_feedback_forward_cosine": 0.9951339960098267 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14897760011672972 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09500788897275925, + "mean_feedback_forward_cosine": 0.9990142898126082, + "mean_feedback_forward_relative_error": 0.04155274192717942, + "min_feedback_forward_cosine": 0.995521605014801 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15326515437602997 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09107612818479538, + "mean_feedback_forward_cosine": 0.9990905924276872, + "mean_feedback_forward_relative_error": 0.03993578402494842, + "min_feedback_forward_cosine": 0.9958833456039429 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14831679260253905 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0873541310429573, + "mean_feedback_forward_cosine": 0.9991615978154269, + "mean_feedback_forward_relative_error": 0.03835834117436951, + "min_feedback_forward_cosine": 0.9962108731269836 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.15207230830192567 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08383745700120926, + "mean_feedback_forward_cosine": 0.9992264563387091, + "mean_feedback_forward_relative_error": 0.036856697042557326, + "min_feedback_forward_cosine": 0.9965096712112427 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15156425470352172 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08054554462432861, + "mean_feedback_forward_cosine": 0.9992848027836193, + "mean_feedback_forward_relative_error": 0.03545508719980717, + "min_feedback_forward_cosine": 0.9967774748802185 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14365550803661348 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0773683413863182, + "mean_feedback_forward_cosine": 0.9993400530381636, + "mean_feedback_forward_relative_error": 0.034076379087160935, + "min_feedback_forward_cosine": 0.997025728225708 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14493907814264298 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07412941008806229, + "mean_feedback_forward_cosine": 0.9993923219767484, + "mean_feedback_forward_relative_error": 0.03272376771677624, + "min_feedback_forward_cosine": 0.9972688555717468 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1477852155828476 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07104866951704025, + "mean_feedback_forward_cosine": 0.999440254948356, + "mean_feedback_forward_relative_error": 0.031422443501651284, + "min_feedback_forward_cosine": 0.997490644454956 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14734023728847503 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06826168298721313, + "mean_feedback_forward_cosine": 0.999483335018158, + "mean_feedback_forward_relative_error": 0.030201662348752673, + "min_feedback_forward_cosine": 0.9976828694343567 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.146364845328331 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0682401955127716, + "mean_feedback_forward_cosine": 0.9994836807250976, + "mean_feedback_forward_relative_error": 0.030190771013836968, + "min_feedback_forward_cosine": 0.9976843595504761 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.0802191784787178 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0682283490896225, + "mean_feedback_forward_cosine": 0.9994838562878695, + "mean_feedback_forward_relative_error": 0.0301847822540863, + "min_feedback_forward_cosine": 0.9976852536201477 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05366209157705307 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06821569800376892, + "mean_feedback_forward_cosine": 0.9994840231808749, + "mean_feedback_forward_relative_error": 0.030179313011467458, + "min_feedback_forward_cosine": 0.9976860880851746 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.043359219493865966 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06820192188024521, + "mean_feedback_forward_cosine": 0.9994842019948093, + "mean_feedback_forward_relative_error": 0.03017386038025672, + "min_feedback_forward_cosine": 0.9976872205734253 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03950208593964577 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06818819791078568, + "mean_feedback_forward_cosine": 0.9994843537157232, + "mean_feedback_forward_relative_error": 0.030168596663596954, + "min_feedback_forward_cosine": 0.9976880550384521 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.034691714227199556 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06817445158958435, + "mean_feedback_forward_cosine": 0.9994845184412869, + "mean_feedback_forward_relative_error": 0.030163414945656605, + "min_feedback_forward_cosine": 0.997688889503479 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.030728123164772986 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06816031783819199, + "mean_feedback_forward_cosine": 0.9994846723296426, + "mean_feedback_forward_relative_error": 0.030158490137281745, + "min_feedback_forward_cosine": 0.9976897835731506 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.02796993443250656 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06814504414796829, + "mean_feedback_forward_cosine": 0.9994848413900895, + "mean_feedback_forward_relative_error": 0.030153311797502365, + "min_feedback_forward_cosine": 0.9976907968521118 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.026594881633855403 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06812984496355057, + "mean_feedback_forward_cosine": 0.9994850147854198, + "mean_feedback_forward_relative_error": 0.030148253691467373, + "min_feedback_forward_cosine": 0.9976917505264282 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.024129524217247962 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0681147351861, + "mean_feedback_forward_cosine": 0.9994851632551713, + "mean_feedback_forward_relative_error": 0.030143321855840358, + "min_feedback_forward_cosine": 0.9976927638053894 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.02287360542178154 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06809647381305695, + "mean_feedback_forward_cosine": 0.9994853464039889, + "mean_feedback_forward_relative_error": 0.030138056649064476, + "min_feedback_forward_cosine": 0.9976940751075745 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.022134124634861946 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06808172911405563, + "mean_feedback_forward_cosine": 0.9994855057109486, + "mean_feedback_forward_relative_error": 0.030133291105316443, + "min_feedback_forward_cosine": 0.9976949095726013 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.019623418084681033 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0680660679936409, + "mean_feedback_forward_cosine": 0.9994856552644209, + "mean_feedback_forward_relative_error": 0.030128194112330677, + "min_feedback_forward_cosine": 0.997696042060852 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.019880822781324387 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06805063039064407, + "mean_feedback_forward_cosine": 0.9994858264923095, + "mean_feedback_forward_relative_error": 0.030123510711233724, + "min_feedback_forward_cosine": 0.997697114944458 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.018001913035809995 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06803379207849503, + "mean_feedback_forward_cosine": 0.9994859890504317, + "mean_feedback_forward_relative_error": 0.030118542156097564, + "min_feedback_forward_cosine": 0.997698187828064 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.016911001521945 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06801775097846985, + "mean_feedback_forward_cosine": 0.9994861570271578, + "mean_feedback_forward_relative_error": 0.030113677976822312, + "min_feedback_forward_cosine": 0.9976992607116699 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.015953069033920766 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06800206005573273, + "mean_feedback_forward_cosine": 0.9994863217527216, + "mean_feedback_forward_relative_error": 0.030109035147523337, + "min_feedback_forward_cosine": 0.9977003931999207 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01511427465081215 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06798448413610458, + "mean_feedback_forward_cosine": 0.9994864918968894, + "mean_feedback_forward_relative_error": 0.03010388956489888, + "min_feedback_forward_cosine": 0.9977014064788818 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.015267606116458774 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06796781718730927, + "mean_feedback_forward_cosine": 0.9994866620410573, + "mean_feedback_forward_relative_error": 0.030099049760875376, + "min_feedback_forward_cosine": 0.9977025985717773 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.014295138305872679 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0679512470960617, + "mean_feedback_forward_cosine": 0.9994868137619712, + "mean_feedback_forward_relative_error": 0.030094342331656, + "min_feedback_forward_cosine": 0.9977036714553833 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.013587301034033298 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06793487071990967, + "mean_feedback_forward_cosine": 0.9994869860735807, + "mean_feedback_forward_relative_error": 0.03008950800550255, + "min_feedback_forward_cosine": 0.9977046847343445 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.012850923472344875 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0679175928235054, + "mean_feedback_forward_cosine": 0.9994871638037942, + "mean_feedback_forward_relative_error": 0.03008446695113724, + "min_feedback_forward_cosine": 0.9977058172225952 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.012713236034810543 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06789865344762802, + "mean_feedback_forward_cosine": 0.9994873437014493, + "mean_feedback_forward_relative_error": 0.030079286494715647, + "min_feedback_forward_cosine": 0.9977070689201355 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.012308287309929729 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06788161396980286, + "mean_feedback_forward_cosine": 0.9994875030084089, + "mean_feedback_forward_relative_error": 0.030074561658230695, + "min_feedback_forward_cosine": 0.9977082014083862 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.011468153971470892 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06786490976810455, + "mean_feedback_forward_cosine": 0.9994876883246682, + "mean_feedback_forward_relative_error": 0.03006964431736957, + "min_feedback_forward_cosine": 0.9977092742919922 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.011168898286223411 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06784682720899582, + "mean_feedback_forward_cosine": 0.9994878628037193, + "mean_feedback_forward_relative_error": 0.030064505321735685, + "min_feedback_forward_cosine": 0.9977104663848877 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.010675356038808823 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06783173233270645, + "mean_feedback_forward_cosine": 0.9994880145246332, + "mean_feedback_forward_relative_error": 0.03006000839681788, + "min_feedback_forward_cosine": 0.9977114796638489 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.009513134524971247 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06781570613384247, + "mean_feedback_forward_cosine": 0.9994881749153137, + "mean_feedback_forward_relative_error": 0.030055251615968618, + "min_feedback_forward_cosine": 0.9977124929428101 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.009519220564886927 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06779937446117401, + "mean_feedback_forward_cosine": 0.9994883407245982, + "mean_feedback_forward_relative_error": 0.030050520903684878, + "min_feedback_forward_cosine": 0.997713565826416 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.009649479085206986 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0677826851606369, + "mean_feedback_forward_cosine": 0.9994885217059742, + "mean_feedback_forward_relative_error": 0.030045833684165368, + "min_feedback_forward_cosine": 0.9977146983146667 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.008831491970270873 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06776418536901474, + "mean_feedback_forward_cosine": 0.9994887016036293, + "mean_feedback_forward_relative_error": 0.03004076196388765, + "min_feedback_forward_cosine": 0.997715950012207 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.009418278399966657 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06774482131004333, + "mean_feedback_forward_cosine": 0.9994888923384927, + "mean_feedback_forward_relative_error": 0.03003532240505923, + "min_feedback_forward_cosine": 0.9977171421051025 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.009592618333250284 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06772809475660324, + "mean_feedback_forward_cosine": 0.9994890505617315, + "mean_feedback_forward_relative_error": 0.030030591946772555, + "min_feedback_forward_cosine": 0.997718334197998 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.007983197383247317 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06770776212215424, + "mean_feedback_forward_cosine": 0.9994892597198486, + "mean_feedback_forward_relative_error": 0.030025187134742736, + "min_feedback_forward_cosine": 0.9977196455001831 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.009227880974523723 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06768783181905746, + "mean_feedback_forward_cosine": 0.9994894396175038, + "mean_feedback_forward_relative_error": 0.030019957292824983, + "min_feedback_forward_cosine": 0.9977210760116577 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.008448275340795517 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06767150014638901, + "mean_feedback_forward_cosine": 0.9994896032593467, + "mean_feedback_forward_relative_error": 0.030015156358819118, + "min_feedback_forward_cosine": 0.9977221488952637 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008099065444543958 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06765541434288025, + "mean_feedback_forward_cosine": 0.9994897788221185, + "mean_feedback_forward_relative_error": 0.030010465922003442, + "min_feedback_forward_cosine": 0.9977231025695801 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.007605297755300998 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06763419508934021, + "mean_feedback_forward_cosine": 0.9994899728081443, + "mean_feedback_forward_relative_error": 0.030004414077848197, + "min_feedback_forward_cosine": 0.9977245330810547 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.008371457745283842 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06761518865823746, + "mean_feedback_forward_cosine": 0.9994901700453325, + "mean_feedback_forward_relative_error": 0.029999074331400068, + "min_feedback_forward_cosine": 0.9977258443832397 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.007859050422012806 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06759700179100037, + "mean_feedback_forward_cosine": 0.9994903640313582, + "mean_feedback_forward_relative_error": 0.02999353918500922, + "min_feedback_forward_cosine": 0.9977270364761353 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.0077766713855415585 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06757808476686478, + "mean_feedback_forward_cosine": 0.9994905764406378, + "mean_feedback_forward_relative_error": 0.02998782027174126, + "min_feedback_forward_cosine": 0.9977282881736755 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.00798575831759721 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06755691766738892, + "mean_feedback_forward_cosine": 0.9994907899336382, + "mean_feedback_forward_relative_error": 0.029981972234831615, + "min_feedback_forward_cosine": 0.9977295994758606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.007603930989652872 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06753963232040405, + "mean_feedback_forward_cosine": 0.9994909600778059, + "mean_feedback_forward_relative_error": 0.02997714238749309, + "min_feedback_forward_cosine": 0.9977308511734009 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.0069355460249260065 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06752147525548935, + "mean_feedback_forward_cosine": 0.9994911237196489, + "mean_feedback_forward_relative_error": 0.029971912274645133, + "min_feedback_forward_cosine": 0.9977318048477173 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006878562457635999 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06750191003084183, + "mean_feedback_forward_cosine": 0.9994913252917204, + "mean_feedback_forward_relative_error": 0.029966585913842374, + "min_feedback_forward_cosine": 0.9977332949638367 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.006689532620459795 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06748202443122864, + "mean_feedback_forward_cosine": 0.9994915268637917, + "mean_feedback_forward_relative_error": 0.029960997360335154, + "min_feedback_forward_cosine": 0.9977346658706665 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006697340769506991 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06746382266283035, + "mean_feedback_forward_cosine": 0.9994917284358632, + "mean_feedback_forward_relative_error": 0.029955772632224993, + "min_feedback_forward_cosine": 0.997735857963562 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.006545494780167937 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06744419038295746, + "mean_feedback_forward_cosine": 0.9994919245893304, + "mean_feedback_forward_relative_error": 0.029950090844861485, + "min_feedback_forward_cosine": 0.9977370500564575 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006664705861844122 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06742171198129654, + "mean_feedback_forward_cosine": 0.999492143500935, + "mean_feedback_forward_relative_error": 0.029943787331946872, + "min_feedback_forward_cosine": 0.9977386593818665 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.00722001444298774 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740182638168335, + "mean_feedback_forward_cosine": 0.9994923548264937, + "mean_feedback_forward_relative_error": 0.02993803441693837, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.006274170103864744 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740158796310425, + "mean_feedback_forward_cosine": 0.9994923656637018, + "mean_feedback_forward_relative_error": 0.029937946440821345, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005793132462501526 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0674014464020729, + "mean_feedback_forward_cosine": 0.9994923548264937, + "mean_feedback_forward_relative_error": 0.029937878962267527, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.005183473689854145 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740128248929977, + "mean_feedback_forward_cosine": 0.999492369998585, + "mean_feedback_forward_relative_error": 0.02993781239810315, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0055938471915386615 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740112602710724, + "mean_feedback_forward_cosine": 0.9994923613288186, + "mean_feedback_forward_relative_error": 0.029937750126489183, + "min_feedback_forward_cosine": 0.9977399706840515 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004966788945272565 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740099936723709, + "mean_feedback_forward_cosine": 0.9994923537427729, + "mean_feedback_forward_relative_error": 0.029937695457854054, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004283378521986306 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740088015794754, + "mean_feedback_forward_cosine": 0.9994923580776561, + "mean_feedback_forward_relative_error": 0.029937641305679625, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004494202756518498 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740075349807739, + "mean_feedback_forward_cosine": 0.9994923580776561, + "mean_feedback_forward_relative_error": 0.029937585951252416, + "min_feedback_forward_cosine": 0.9977399110794067 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004588802584782243 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740062683820724, + "mean_feedback_forward_cosine": 0.9994923765009099, + "mean_feedback_forward_relative_error": 0.02993752999569882, + "min_feedback_forward_cosine": 0.9977399706840515 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.004372061618436128 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740052253007889, + "mean_feedback_forward_cosine": 0.9994923743334684, + "mean_feedback_forward_relative_error": 0.029937479526481844, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0039116406575217845 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740038096904755, + "mean_feedback_forward_cosine": 0.9994923656637018, + "mean_feedback_forward_relative_error": 0.02993742266500538, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.00434201705057174 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0674002394080162, + "mean_feedback_forward_cosine": 0.9994923808357933, + "mean_feedback_forward_relative_error": 0.029937365227802232, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0043814171719551084 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06740012019872665, + "mean_feedback_forward_cosine": 0.999492369998585, + "mean_feedback_forward_relative_error": 0.02993730908598412, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.004154835649877787 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0674000158905983, + "mean_feedback_forward_cosine": 0.9994923765009099, + "mean_feedback_forward_relative_error": 0.029937259150160985, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.003897070373259485 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739989668130875, + "mean_feedback_forward_cosine": 0.9994923569939353, + "mean_feedback_forward_relative_error": 0.029937205649912358, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.004187560288645327 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673997774720192, + "mean_feedback_forward_cosine": 0.9994923754171892, + "mean_feedback_forward_relative_error": 0.029937153445048766, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0038901928127743303 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739966571331024, + "mean_feedback_forward_cosine": 0.9994923830032348, + "mean_feedback_forward_relative_error": 0.029937104101885447, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003830749564897269 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673995390534401, + "mean_feedback_forward_cosine": 0.9994923786683516, + "mean_feedback_forward_relative_error": 0.029937052735212175, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0034973140881955625 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739944219589233, + "mean_feedback_forward_cosine": 0.9994923797520724, + "mean_feedback_forward_relative_error": 0.029937005034563216, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.00363929495383054 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739933788776398, + "mean_feedback_forward_cosine": 0.9994924014264887, + "mean_feedback_forward_relative_error": 0.029936956089328635, + "min_feedback_forward_cosine": 0.9977400302886963 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0038257140539586544 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739923357963562, + "mean_feedback_forward_cosine": 0.9994923830032348, + "mean_feedback_forward_relative_error": 0.02993690621277148, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0037405504473671317 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739912927150726, + "mean_feedback_forward_cosine": 0.9994923754171892, + "mean_feedback_forward_relative_error": 0.029936855489557438, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0036961491615790875 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673990324139595, + "mean_feedback_forward_cosine": 0.9994923743334684, + "mean_feedback_forward_relative_error": 0.029936808796430178, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003587280821725726 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739892810583115, + "mean_feedback_forward_cosine": 0.9994923765009099, + "mean_feedback_forward_relative_error": 0.029936761358244853, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.003504292621985078 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739881634712219, + "mean_feedback_forward_cosine": 0.9994923851706765, + "mean_feedback_forward_relative_error": 0.02993671512231231, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.003447736492715776 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739870458841324, + "mean_feedback_forward_cosine": 0.9994923689148643, + "mean_feedback_forward_relative_error": 0.029936667091467164, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.003741029561981559 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739861518144608, + "mean_feedback_forward_cosine": 0.9994923895055597, + "mean_feedback_forward_relative_error": 0.02993662009354342, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.003317566669508815 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739848852157593, + "mean_feedback_forward_cosine": 0.9994923992590471, + "mean_feedback_forward_relative_error": 0.029936568532139062, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.003686123422048986 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739836186170578, + "mean_feedback_forward_cosine": 0.9994923819195141, + "mean_feedback_forward_relative_error": 0.02993651358410716, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.003723716419087723 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739827245473862, + "mean_feedback_forward_cosine": 0.9994923981753263, + "mean_feedback_forward_relative_error": 0.029936465468596327, + "min_feedback_forward_cosine": 0.9977400898933411 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0032948575318977238 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739818304777145, + "mean_feedback_forward_cosine": 0.9994923927567222, + "mean_feedback_forward_relative_error": 0.029936423313549974, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.003050474014468491 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673980712890625, + "mean_feedback_forward_cosine": 0.9994923960078846, + "mean_feedback_forward_relative_error": 0.029936372802000153, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0038788184327445924 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739798188209534, + "mean_feedback_forward_cosine": 0.9994923862543973, + "mean_feedback_forward_relative_error": 0.029936325973407788, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0031438035936653614 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739788502454758, + "mean_feedback_forward_cosine": 0.9994923992590471, + "mean_feedback_forward_relative_error": 0.029936281659386374, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0033135824465192853 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739780306816101, + "mean_feedback_forward_cosine": 0.9994923808357933, + "mean_feedback_forward_relative_error": 0.029936238877813926, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.00299471766974777 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739768385887146, + "mean_feedback_forward_cosine": 0.999492388421839, + "mean_feedback_forward_relative_error": 0.029936187942935663, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0033631774935498834 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673975944519043, + "mean_feedback_forward_cosine": 0.9994923851706765, + "mean_feedback_forward_relative_error": 0.02993614422157407, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0030226892620697616 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739750504493713, + "mean_feedback_forward_cosine": 0.9994923949241639, + "mean_feedback_forward_relative_error": 0.029936096859587865, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.003323344582170248 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739738583564758, + "mean_feedback_forward_cosine": 0.9994924025102095, + "mean_feedback_forward_relative_error": 0.029936048735610465, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003273675970211625 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739726662635803, + "mean_feedback_forward_cosine": 0.9994923992590471, + "mean_feedback_forward_relative_error": 0.029936002347279677, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003436235188674182 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739716976881027, + "mean_feedback_forward_cosine": 0.9994924003427679, + "mean_feedback_forward_relative_error": 0.029935957135801966, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0031318929681554437 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739707291126251, + "mean_feedback_forward_cosine": 0.9994923927567222, + "mean_feedback_forward_relative_error": 0.029935914345762947, + "min_feedback_forward_cosine": 0.9977401494979858 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.003197696532420814 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739697605371475, + "mean_feedback_forward_cosine": 0.9994923960078846, + "mean_feedback_forward_relative_error": 0.029935869252817198, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0032609035214222967 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739687919616699, + "mean_feedback_forward_cosine": 0.9994924176823009, + "mean_feedback_forward_relative_error": 0.029935823677277023, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0031454689562600106 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739677488803864, + "mean_feedback_forward_cosine": 0.9994924133474177, + "mean_feedback_forward_relative_error": 0.029935773572122508, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.003257605445832014 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739669293165207, + "mean_feedback_forward_cosine": 0.9994923970916054, + "mean_feedback_forward_relative_error": 0.02993572960523042, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0028618731648474933 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0673966109752655, + "mean_feedback_forward_cosine": 0.9994924133474177, + "mean_feedback_forward_relative_error": 0.029935687831179662, + "min_feedback_forward_cosine": 0.9977402091026306 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.002991606337893754 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739653646945953, + "mean_feedback_forward_cosine": 0.9994923992590471, + "mean_feedback_forward_relative_error": 0.029935647928240626, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0027888119074143468 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739643961191177, + "mean_feedback_forward_cosine": 0.9994924133474177, + "mean_feedback_forward_relative_error": 0.029935601328245617, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0034980247987434266 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739633530378342, + "mean_feedback_forward_cosine": 0.9994924187660217, + "mean_feedback_forward_relative_error": 0.029935555456375534, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.003266522671226412 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06739622354507446, + "mean_feedback_forward_cosine": 0.9994924046776511, + "mean_feedback_forward_relative_error": 0.0299355069598691, + "min_feedback_forward_cosine": 0.9977402687072754 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0032272437416017057 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9273, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.390718896484375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3318545920, + "peak_memory_reserved_bytes": 4754243584, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0538651943206787, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 18248.176662921906, + "train_wall_s": 18242.354815006256 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250433280000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 5010580480000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json new file mode 100644 index 0000000..9ed100a --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json @@ -0,0 +1,3654 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9993968506654104, + "feedback_forward_cosine": [ + 0.9995486736297607, + 0.9991993308067322, + 0.9991708993911743, + 0.9989316463470459, + 0.9989402294158936, + 0.998775839805603, + 0.9987432956695557, + 0.9984483122825623, + 0.999407172203064, + 0.9991537928581238, + 0.9992674589157104, + 0.9991320371627808, + 0.9986305236816406, + 0.9982662796974182, + 0.9992818236351013, + 0.9991592764854431, + 0.9995694756507874, + 0.9995255470275879, + 0.9996371865272522, + 0.9996630549430847, + 0.9987449645996094, + 0.9988316297531128, + 0.9991240501403809, + 0.999101996421814, + 0.9987897872924805, + 0.9986447691917419, + 0.9992947578430176, + 0.999209463596344, + 0.9993540048599243, + 0.9992989301681519, + 0.9994040131568909, + 0.999302089214325, + 0.9996514916419983, + 0.999592661857605, + 0.9998459219932556, + 0.9997779130935669, + 0.9997722506523132, + 0.9998602271080017, + 0.9997929334640503, + 0.9997841119766235, + 0.9997772574424744, + 0.9997502565383911, + 0.9997769594192505, + 0.9997308850288391, + 0.9997909069061279, + 0.999720573425293, + 0.9997861385345459, + 0.9996883869171143, + 0.9997413158416748, + 0.9996073246002197, + 0.9997631907463074, + 0.99956876039505, + 0.9996873140335083, + 0.9993549585342407, + 0.9999806880950928 + ], + "feedback_forward_norm_ratio": [ + 0.999349057674408, + 1.0004899501800537, + 0.9993137121200562, + 1.0019783973693848, + 0.9970383644104004, + 1.001090168952942, + 0.9982687830924988, + 1.0007696151733398, + 0.9992757439613342, + 1.0008498430252075, + 0.9987574219703674, + 1.00188410282135, + 1.002251148223877, + 1.0031952857971191, + 0.9994503259658813, + 0.9994155168533325, + 0.9998971819877625, + 1.001420497894287, + 0.9988576769828796, + 1.000478982925415, + 0.9980266690254211, + 1.001484990119934, + 0.9985701441764832, + 1.0008782148361206, + 0.9986444115638733, + 1.0008747577667236, + 0.9995606541633606, + 1.0017122030258179, + 0.9987946152687073, + 1.0015406608581543, + 0.9989224672317505, + 1.0014458894729614, + 0.9994480609893799, + 1.0008728504180908, + 0.9996688961982727, + 1.0007503032684326, + 0.9992460012435913, + 1.0002858638763428, + 0.9995903372764587, + 1.0005089044570923, + 0.9994708895683289, + 1.0008928775787354, + 0.9995566606521606, + 1.0010942220687866, + 0.9995842576026917, + 1.0011290311813354, + 0.9995908141136169, + 1.0012075901031494, + 0.9993408918380737, + 1.0018969774246216, + 0.9994351267814636, + 1.0011729001998901, + 0.9998877048492432, + 1.0017391443252563, + 1.0021624565124512 + ], + "feedback_forward_relative_error": [ + 0.03004281036555767, + 0.04002983868122101, + 0.04071551933884621, + 0.04631281644105911, + 0.04606624320149422, + 0.0495198555290699, + 0.05012083426117897, + 0.05573458969593048, + 0.034428082406520844, + 0.041166696697473526, + 0.03827217221260071, + 0.0417473204433918, + 0.05244080722332001, + 0.059065550565719604, + 0.03788788989186287, + 0.04099579527974129, + 0.029339220374822617, + 0.030859628692269325, + 0.02694813348352909, + 0.02597028762102127, + 0.050091758370399475, + 0.04839812591671944, + 0.04184954985976219, + 0.04240509867668152, + 0.04918261989951134, + 0.052094217389822006, + 0.03755047917366028, + 0.03983255475759506, + 0.03594253957271576, + 0.0375065915286541, + 0.03452565148472786, + 0.03741465136408806, + 0.026397136971354485, + 0.028567994013428688, + 0.017552809789776802, + 0.02109622210264206, + 0.021345283836126328, + 0.016726097092032433, + 0.020349428057670593, + 0.020791415125131607, + 0.021108200773596764, + 0.022380851209163666, + 0.02111971750855446, + 0.023237867280840874, + 0.02044793963432312, + 0.023680970072746277, + 0.02068115770816803, + 0.025010209530591965, + 0.022751644253730774, + 0.028116144239902496, + 0.021761473268270493, + 0.02940809354186058, + 0.02500557154417038, + 0.03599012270569801, + 0.006583130918443203 + ], + "innovation_negative_gradient_cosine": [ + 0.9995798468589783, + 0.9994189739227295, + 0.9995826482772827, + 0.999254584312439, + 0.9995844960212708, + 0.9990823268890381, + 0.9995860457420349, + 0.9990700483322144, + 0.9995884299278259, + 0.9992920160293579, + 0.9995938539505005, + 0.9992960691452026, + 0.9995992183685303, + 0.9988152384757996, + 0.9995999336242676, + 0.9992086887359619, + 0.9996014833450317, + 0.9993894100189209, + 0.9996145367622375, + 0.999541163444519, + 0.9996229410171509, + 0.9991058111190796, + 0.9996281862258911, + 0.9992707371711731, + 0.999628484249115, + 0.9991620779037476, + 0.9996294975280762, + 0.9992156624794006, + 0.9996325969696045, + 0.9993586540222168, + 0.9996317625045776, + 0.9994008541107178, + 0.9996525049209595, + 0.9994752407073975, + 0.9996588230133057, + 0.9996336102485657, + 0.9996620416641235, + 0.9996757507324219, + 0.9997068047523499, + 0.9996463656425476, + 0.999724805355072, + 0.9996492862701416, + 0.9997373819351196, + 0.9996373653411865, + 0.9997483491897583, + 0.9996845722198486, + 0.9997813701629639, + 0.999725878238678, + 0.9998012781143188, + 0.999775767326355, + 0.9998584985733032, + 0.9998941421508789, + 0.9999178647994995, + 0.9999518394470215, + 0.8171511888504028 + ], + "mean_feedback_forward_cosine": 0.9993700861930848, + "mean_feedback_forward_relative_error": 0.03353758930143985, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9995798468589783, + 0.9994189739227295, + 0.9995826482772827, + 0.999254584312439, + 0.9995844960212708, + 0.9990823268890381, + 0.9995860457420349, + 0.9990700483322144, + 0.9995884299278259, + 0.9992920160293579, + 0.9995938539505005, + 0.9992960691452026, + 0.9995992183685303, + 0.9988152384757996, + 0.9995999336242676, + 0.9992086887359619, + 0.9996014833450317, + 0.9993894100189209, + 0.9996145367622375, + 0.999541163444519, + 0.9996229410171509, + 0.9991058111190796, + 0.9996281862258911, + 0.9992707371711731, + 0.999628484249115, + 0.9991620779037476, + 0.9996294975280762, + 0.9992156624794006, + 0.9996325969696045, + 0.9993586540222168, + 0.9996317625045776, + 0.9994008541107178, + 0.9996525049209595, + 0.9994752407073975, + 0.9996588230133057, + 0.9996336102485657, + 0.9996620416641235, + 0.9996757507324219, + 0.9997068047523499, + 0.9996463656425476, + 0.999724805355072, + 0.9996492862701416, + 0.9997373819351196, + 0.9996373653411865, + 0.9997483491897583, + 0.9996845722198486, + 0.9997813701629639, + 0.999725878238678, + 0.9998012781143188, + 0.999775767326355, + 0.9998584985733032, + 0.9998941421508789, + 0.9999178647994995, + 0.9999518394470215, + 0.8171511888504028 + ], + "teaching_negative_gradient_cosine": [ + 0.9995798468589783, + 0.9994189739227295, + 0.9995826482772827, + 0.999254584312439, + 0.9995844960212708, + 0.9990823268890381, + 0.9995860457420349, + 0.9990700483322144, + 0.9995884299278259, + 0.9992920160293579, + 0.9995938539505005, + 0.9992960691452026, + 0.9995992183685303, + 0.9988152384757996, + 0.9995999336242676, + 0.9992086887359619, + 0.9996014833450317, + 0.9993894100189209, + 0.9996145367622375, + 0.999541163444519, + 0.9996229410171509, + 0.9991058111190796, + 0.9996281862258911, + 0.9992707371711731, + 0.999628484249115, + 0.9991620779037476, + 0.9996294975280762, + 0.9992156624794006, + 0.9996325969696045, + 0.9993586540222168, + 0.9996317625045776, + 0.9994008541107178, + 0.9996525049209595, + 0.9994752407073975, + 0.9996588230133057, + 0.9996336102485657, + 0.9996620416641235, + 0.9996757507324219, + 0.9997068047523499, + 0.9996463656425476, + 0.999724805355072, + 0.9996492862701416, + 0.9997373819351196, + 0.9996373653411865, + 0.9997483491897583, + 0.9996845722198486, + 0.9997813701629639, + 0.999725878238678, + 0.9998012781143188, + 0.999775767326355, + 0.9998584985733032, + 0.9998941421508789, + 0.9999178647994995, + 0.9999518394470215, + 0.8171511888504028 + ], + "wall_s": 0.17903757095336914 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3360395431518555, + "mean_feedback_forward_cosine": 0.5398452989079735, + "mean_feedback_forward_relative_error": 0.9057000404054468, + "min_feedback_forward_cosine": 0.11729566752910614 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.261840078430176 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.301427960395813, + "mean_feedback_forward_cosine": 0.5570171255956996, + "mean_feedback_forward_relative_error": 0.8909265252676877, + "min_feedback_forward_cosine": 0.1651471108198166 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.632542042617798 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2671263217926025, + "mean_feedback_forward_cosine": 0.5770768756216222, + "mean_feedback_forward_relative_error": 0.8736525118350983, + "min_feedback_forward_cosine": 0.2004789412021637 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.43858590965271 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2385221719741821, + "mean_feedback_forward_cosine": 0.5996637119488283, + "mean_feedback_forward_relative_error": 0.8534463828260248, + "min_feedback_forward_cosine": 0.23759151995182037 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.2779202951812745 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2044333219528198, + "mean_feedback_forward_cosine": 0.6234834736043756, + "mean_feedback_forward_relative_error": 0.8310065833005038, + "min_feedback_forward_cosine": 0.28002744913101196 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.125606082649231 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.168691873550415, + "mean_feedback_forward_cosine": 0.6454366548494859, + "mean_feedback_forward_relative_error": 0.8094265899874947, + "min_feedback_forward_cosine": 0.3218994140625 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 1.001765019683838 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1300506591796875, + "mean_feedback_forward_cosine": 0.6663690940900282, + "mean_feedback_forward_relative_error": 0.7878504964438352, + "min_feedback_forward_cosine": 0.36677074432373047 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.8995718880844116 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0915017127990723, + "mean_feedback_forward_cosine": 0.6862029666250402, + "mean_feedback_forward_relative_error": 0.7663991510868072, + "min_feedback_forward_cosine": 0.40944093465805054 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.8154410832023621 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.062315583229065, + "mean_feedback_forward_cosine": 0.7044076865369623, + "mean_feedback_forward_relative_error": 0.7458936360749331, + "min_feedback_forward_cosine": 0.45194125175476074 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.7438993234443665 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0368866920471191, + "mean_feedback_forward_cosine": 0.7206309345635501, + "mean_feedback_forward_relative_error": 0.7267518672076139, + "min_feedback_forward_cosine": 0.482383131980896 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.6833764509677887 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.01106858253479, + "mean_feedback_forward_cosine": 0.7355885321443731, + "mean_feedback_forward_relative_error": 0.7083800781856884, + "min_feedback_forward_cosine": 0.5088209509849548 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.6408538368225097 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9844040274620056, + "mean_feedback_forward_cosine": 0.7500098434361544, + "mean_feedback_forward_relative_error": 0.690076062896035, + "min_feedback_forward_cosine": 0.5345862507820129 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.6051172892189026 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9599705338478088, + "mean_feedback_forward_cosine": 0.7629830067807978, + "mean_feedback_forward_relative_error": 0.6729642987251282, + "min_feedback_forward_cosine": 0.5579694509506226 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.5670238526153565 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9349731206893921, + "mean_feedback_forward_cosine": 0.7752044851129706, + "mean_feedback_forward_relative_error": 0.656314297968691, + "min_feedback_forward_cosine": 0.5811662077903748 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.5346621171951293 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9106687307357788, + "mean_feedback_forward_cosine": 0.7867725946686485, + "mean_feedback_forward_relative_error": 0.6399037128145044, + "min_feedback_forward_cosine": 0.6029323935508728 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.5101796304035187 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8859759569168091, + "mean_feedback_forward_cosine": 0.7978461677377874, + "mean_feedback_forward_relative_error": 0.6236773769963871, + "min_feedback_forward_cosine": 0.6241481304168701 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.49010099882125857 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8601457476615906, + "mean_feedback_forward_cosine": 0.8083188620480624, + "mean_feedback_forward_relative_error": 0.6077677271582863, + "min_feedback_forward_cosine": 0.6459009647369385 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.4727135664749145 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8357653617858887, + "mean_feedback_forward_cosine": 0.8180293473330411, + "mean_feedback_forward_relative_error": 0.592511474815282, + "min_feedback_forward_cosine": 0.6656611561775208 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.45012226643562314 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8113493919372559, + "mean_feedback_forward_cosine": 0.8274970444765958, + "mean_feedback_forward_relative_error": 0.5771354071118615, + "min_feedback_forward_cosine": 0.6852105855941772 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.4389621475124359 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7879949808120728, + "mean_feedback_forward_cosine": 0.8362948872826316, + "mean_feedback_forward_relative_error": 0.5622997183691372, + "min_feedback_forward_cosine": 0.7030802369117737 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.4171404622268677 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7637096047401428, + "mean_feedback_forward_cosine": 0.8449254285205494, + "mean_feedback_forward_relative_error": 0.5472864966500889, + "min_feedback_forward_cosine": 0.7209082841873169 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.4103449504566193 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7398356795310974, + "mean_feedback_forward_cosine": 0.8530320265076378, + "mean_feedback_forward_relative_error": 0.5328117982907729, + "min_feedback_forward_cosine": 0.7381588220596313 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3970993436813354 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.716636598110199, + "mean_feedback_forward_cosine": 0.8606991063464772, + "mean_feedback_forward_relative_error": 0.5185809793797407, + "min_feedback_forward_cosine": 0.75416499376297 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3831087645149231 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6941056251525879, + "mean_feedback_forward_cosine": 0.8678555889563127, + "mean_feedback_forward_relative_error": 0.5049192954193462, + "min_feedback_forward_cosine": 0.7692728042602539 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.36635501386642455 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6716936230659485, + "mean_feedback_forward_cosine": 0.8749791730533947, + "mean_feedback_forward_relative_error": 0.49094241992993787, + "min_feedback_forward_cosine": 0.7837886810302734 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3610792467689514 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6493027210235596, + "mean_feedback_forward_cosine": 0.8815253160216592, + "mean_feedback_forward_relative_error": 0.47770702784711666, + "min_feedback_forward_cosine": 0.7960513234138489 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.3519402063179016 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6338154673576355, + "mean_feedback_forward_cosine": 0.8878368030894886, + "mean_feedback_forward_relative_error": 0.4645320017229427, + "min_feedback_forward_cosine": 0.8039942979812622 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.34612735505104064 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6198217272758484, + "mean_feedback_forward_cosine": 0.8939330339431762, + "mean_feedback_forward_relative_error": 0.4514224285429174, + "min_feedback_forward_cosine": 0.8125402331352234 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.3347622646522522 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6074491739273071, + "mean_feedback_forward_cosine": 0.8996492028236389, + "mean_feedback_forward_relative_error": 0.43872043680060996, + "min_feedback_forward_cosine": 0.8199578523635864 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.327846246509552 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5957520604133606, + "mean_feedback_forward_cosine": 0.9050342408093539, + "mean_feedback_forward_relative_error": 0.42643835246562956, + "min_feedback_forward_cosine": 0.8265140652656555 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.31661326223373415 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5830815434455872, + "mean_feedback_forward_cosine": 0.9101586894555526, + "mean_feedback_forward_relative_error": 0.4143924686041745, + "min_feedback_forward_cosine": 0.8340471386909485 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.303727802362442 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5698826313018799, + "mean_feedback_forward_cosine": 0.9152525284073569, + "mean_feedback_forward_relative_error": 0.40209951116280124, + "min_feedback_forward_cosine": 0.8409469127655029 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.30891136761665344 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5572258830070496, + "mean_feedback_forward_cosine": 0.9199291597713124, + "mean_feedback_forward_relative_error": 0.3904243734749881, + "min_feedback_forward_cosine": 0.8477466106414795 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2947652544689178 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5451564788818359, + "mean_feedback_forward_cosine": 0.9245218396186828, + "mean_feedback_forward_relative_error": 0.37870549031279305, + "min_feedback_forward_cosine": 0.8543558120727539 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.29327453715324403 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5320948958396912, + "mean_feedback_forward_cosine": 0.928714085708965, + "mean_feedback_forward_relative_error": 0.36763537986712025, + "min_feedback_forward_cosine": 0.8603912591934204 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.28498666655540467 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5205531716346741, + "mean_feedback_forward_cosine": 0.9328078356656161, + "mean_feedback_forward_relative_error": 0.35648485327308826, + "min_feedback_forward_cosine": 0.8662968873977661 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.27884571946144104 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5085384249687195, + "mean_feedback_forward_cosine": 0.9367399844256314, + "mean_feedback_forward_relative_error": 0.34547765525904567, + "min_feedback_forward_cosine": 0.8721722960472107 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2728859224414825 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4989911913871765, + "mean_feedback_forward_cosine": 0.9404762939973311, + "mean_feedback_forward_relative_error": 0.33477059453725816, + "min_feedback_forward_cosine": 0.8767096996307373 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2726372246932983 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4896278977394104, + "mean_feedback_forward_cosine": 0.9438534920865839, + "mean_feedback_forward_relative_error": 0.3247550377791578, + "min_feedback_forward_cosine": 0.8813626766204834 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2597659772014618 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4794587194919586, + "mean_feedback_forward_cosine": 0.9471759416840293, + "mean_feedback_forward_relative_error": 0.31466083838181064, + "min_feedback_forward_cosine": 0.8863492608070374 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2566597058677673 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46974116563796997, + "mean_feedback_forward_cosine": 0.9502717776732011, + "mean_feedback_forward_relative_error": 0.3049257840622555, + "min_feedback_forward_cosine": 0.8909269571304321 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2535866453552246 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45840635895729065, + "mean_feedback_forward_cosine": 0.953336640921506, + "mean_feedback_forward_relative_error": 0.29503738731145857, + "min_feedback_forward_cosine": 0.8961610794067383 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2539304351520538 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4476323127746582, + "mean_feedback_forward_cosine": 0.956215010989796, + "mean_feedback_forward_relative_error": 0.2854661605574868, + "min_feedback_forward_cosine": 0.9009397625923157 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.24867324002742766 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4384644627571106, + "mean_feedback_forward_cosine": 0.9589088060639122, + "mean_feedback_forward_relative_error": 0.276235669715838, + "min_feedback_forward_cosine": 0.9049193263053894 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.24410620104789735 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42801737785339355, + "mean_feedback_forward_cosine": 0.9613754695112054, + "mean_feedback_forward_relative_error": 0.2675473301248117, + "min_feedback_forward_cosine": 0.9095538854598999 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.23810331386089326 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41825419664382935, + "mean_feedback_forward_cosine": 0.9637145074931058, + "mean_feedback_forward_relative_error": 0.2590205722234466, + "min_feedback_forward_cosine": 0.9137883186340332 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.23533988403320313 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4078068435192108, + "mean_feedback_forward_cosine": 0.9660646536133506, + "mean_feedback_forward_relative_error": 0.2501983148130504, + "min_feedback_forward_cosine": 0.9179442524909973 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.23375252881526948 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3972949683666229, + "mean_feedback_forward_cosine": 0.9681946667757901, + "mean_feedback_forward_relative_error": 0.24198557592251085, + "min_feedback_forward_cosine": 0.9221304059028625 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.22916575040340423 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3864702880382538, + "mean_feedback_forward_cosine": 0.9702675407583063, + "mean_feedback_forward_relative_error": 0.23377114974639632, + "min_feedback_forward_cosine": 0.9264482259750366 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2279023641014099 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3768327236175537, + "mean_feedback_forward_cosine": 0.9721953045238149, + "mean_feedback_forward_relative_error": 0.22584167129614136, + "min_feedback_forward_cosine": 0.9300549626350403 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.22254173853874207 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3676289916038513, + "mean_feedback_forward_cosine": 0.9739825508811257, + "mean_feedback_forward_relative_error": 0.21825949299064548, + "min_feedback_forward_cosine": 0.9334088563919067 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.216740937871933 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3589218854904175, + "mean_feedback_forward_cosine": 0.9756713390350342, + "mean_feedback_forward_relative_error": 0.2108152225613594, + "min_feedback_forward_cosine": 0.9364010095596313 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.21592123507976532 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34930217266082764, + "mean_feedback_forward_cosine": 0.97728990316391, + "mean_feedback_forward_relative_error": 0.20350951477885246, + "min_feedback_forward_cosine": 0.939773440361023 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.21098066742420196 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3383598029613495, + "mean_feedback_forward_cosine": 0.9788350506262345, + "mean_feedback_forward_relative_error": 0.19632480801506477, + "min_feedback_forward_cosine": 0.9434282779693604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.21545665407657624 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3280719518661499, + "mean_feedback_forward_cosine": 0.9802916006608443, + "mean_feedback_forward_relative_error": 0.18931772085753354, + "min_feedback_forward_cosine": 0.9468652009963989 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.2137152345609665 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3173877000808716, + "mean_feedback_forward_cosine": 0.9816158673980019, + "mean_feedback_forward_relative_error": 0.182749734277075, + "min_feedback_forward_cosine": 0.9502993226051331 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.2024301117658615 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3077414333820343, + "mean_feedback_forward_cosine": 0.9828643029386347, + "mean_feedback_forward_relative_error": 0.17626704580404542, + "min_feedback_forward_cosine": 0.9531734585762024 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.2101277583503723 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29863837361335754, + "mean_feedback_forward_cosine": 0.9840318376367743, + "mean_feedback_forward_relative_error": 0.1699988741766323, + "min_feedback_forward_cosine": 0.9558677673339844 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.20592128714084626 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28895822167396545, + "mean_feedback_forward_cosine": 0.985127288644964, + "mean_feedback_forward_relative_error": 0.16398801268501714, + "min_feedback_forward_cosine": 0.9586999416351318 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.20172549364566802 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2780231237411499, + "mean_feedback_forward_cosine": 0.9861920714378357, + "mean_feedback_forward_relative_error": 0.15795523747801782, + "min_feedback_forward_cosine": 0.9618793725967407 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.20062646031856537 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26860901713371277, + "mean_feedback_forward_cosine": 0.9871679197658192, + "mean_feedback_forward_relative_error": 0.1521740548312664, + "min_feedback_forward_cosine": 0.9643744826316833 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1979875206375122 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25864025950431824, + "mean_feedback_forward_cosine": 0.9880671089345758, + "mean_feedback_forward_relative_error": 0.14666806825182654, + "min_feedback_forward_cosine": 0.9669016599655151 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19849038482666015 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24903854727745056, + "mean_feedback_forward_cosine": 0.9888936259529807, + "mean_feedback_forward_relative_error": 0.1414306450296532, + "min_feedback_forward_cosine": 0.9693467617034912 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.19536169333457948 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2402242124080658, + "mean_feedback_forward_cosine": 0.9896845037286932, + "mean_feedback_forward_relative_error": 0.13624327758496457, + "min_feedback_forward_cosine": 0.9714638590812683 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.19034876211166382 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23262028396129608, + "mean_feedback_forward_cosine": 0.9904228037053888, + "mean_feedback_forward_relative_error": 0.13119625354355033, + "min_feedback_forward_cosine": 0.9732394218444824 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.19610364086151122 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22418992221355438, + "mean_feedback_forward_cosine": 0.9911059455438094, + "mean_feedback_forward_relative_error": 0.12638776359910314, + "min_feedback_forward_cosine": 0.9751158952713013 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18961908777713776 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21725112199783325, + "mean_feedback_forward_cosine": 0.9917246786030856, + "mean_feedback_forward_relative_error": 0.12187630513852293, + "min_feedback_forward_cosine": 0.9766388535499573 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18855149097442628 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21030044555664062, + "mean_feedback_forward_cosine": 0.9923040400851857, + "mean_feedback_forward_relative_error": 0.11746241463856263, + "min_feedback_forward_cosine": 0.9781090021133423 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1827023038291931 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20252923667430878, + "mean_feedback_forward_cosine": 0.9928736892613498, + "mean_feedback_forward_relative_error": 0.11299660514024171, + "min_feedback_forward_cosine": 0.9796693921089172 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18777125893592833 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19475862383842468, + "mean_feedback_forward_cosine": 0.9933944236148488, + "mean_feedback_forward_relative_error": 0.1087731986221942, + "min_feedback_forward_cosine": 0.9812123775482178 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1827207085800171 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18700455129146576, + "mean_feedback_forward_cosine": 0.993892765045166, + "mean_feedback_forward_relative_error": 0.10458857572891496, + "min_feedback_forward_cosine": 0.9826465845108032 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18099756442070009 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18052984774112701, + "mean_feedback_forward_cosine": 0.9943497278473594, + "mean_feedback_forward_relative_error": 0.1005988916551525, + "min_feedback_forward_cosine": 0.9838318228721619 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.180845529088974 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17324385046958923, + "mean_feedback_forward_cosine": 0.9947746981273998, + "mean_feedback_forward_relative_error": 0.09673330394381827, + "min_feedback_forward_cosine": 0.9851095676422119 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18006097470283508 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1666959673166275, + "mean_feedback_forward_cosine": 0.9951619343324141, + "mean_feedback_forward_relative_error": 0.09307177717035467, + "min_feedback_forward_cosine": 0.9862115979194641 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.1777750191783905 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16139131784439087, + "mean_feedback_forward_cosine": 0.9955085570161993, + "mean_feedback_forward_relative_error": 0.08965976339849559, + "min_feedback_forward_cosine": 0.987093448638916 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1749332488489151 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15509504079818726, + "mean_feedback_forward_cosine": 0.9958382075483149, + "mean_feedback_forward_relative_error": 0.08627855002202771, + "min_feedback_forward_cosine": 0.9880722165107727 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17808644639015198 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1499006450176239, + "mean_feedback_forward_cosine": 0.9961370966651223, + "mean_feedback_forward_relative_error": 0.08310288553210822, + "min_feedback_forward_cosine": 0.9888517260551453 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17081560897827147 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1446233093738556, + "mean_feedback_forward_cosine": 0.9964214520020919, + "mean_feedback_forward_relative_error": 0.07997748160904104, + "min_feedback_forward_cosine": 0.9896301627159119 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1707915583229065 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13814088702201843, + "mean_feedback_forward_cosine": 0.996705302325162, + "mean_feedback_forward_relative_error": 0.07675201486457478, + "min_feedback_forward_cosine": 0.990538477897644 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17181168204307556 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13184763491153717, + "mean_feedback_forward_cosine": 0.9969579989259894, + "mean_feedback_forward_relative_error": 0.07376226691359823, + "min_feedback_forward_cosine": 0.9913761615753174 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.16899568368911744 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12670230865478516, + "mean_feedback_forward_cosine": 0.9971872037107294, + "mean_feedback_forward_relative_error": 0.07091320035132495, + "min_feedback_forward_cosine": 0.9920350313186646 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17493566308021546 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12146414816379547, + "mean_feedback_forward_cosine": 0.9973919738422741, + "mean_feedback_forward_relative_error": 0.06827812296423046, + "min_feedback_forward_cosine": 0.9926751255989075 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1669038712310791 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11693733185529709, + "mean_feedback_forward_cosine": 0.9975873134352944, + "mean_feedback_forward_relative_error": 0.06566752669486133, + "min_feedback_forward_cosine": 0.9932080507278442 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.1658772466468811 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11253160238265991, + "mean_feedback_forward_cosine": 0.9977676835927096, + "mean_feedback_forward_relative_error": 0.06316453343765302, + "min_feedback_forward_cosine": 0.9937189817428589 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1679341847229004 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10800867527723312, + "mean_feedback_forward_cosine": 0.99793504151431, + "mean_feedback_forward_relative_error": 0.06075672393834049, + "min_feedback_forward_cosine": 0.9942063093185425 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16437279180526734 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10382198542356491, + "mean_feedback_forward_cosine": 0.9980900504372336, + "mean_feedback_forward_relative_error": 0.058428227579729125, + "min_feedback_forward_cosine": 0.9946441650390625 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16653217509269713 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10005079954862595, + "mean_feedback_forward_cosine": 0.9982323559847746, + "mean_feedback_forward_relative_error": 0.05619608253579248, + "min_feedback_forward_cosine": 0.9950335621833801 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1683104187965393 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09619463235139847, + "mean_feedback_forward_cosine": 0.9983689481561834, + "mean_feedback_forward_relative_error": 0.053986180065707726, + "min_feedback_forward_cosine": 0.9954037070274353 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16667857217788695 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09265120327472687, + "mean_feedback_forward_cosine": 0.998490385575728, + "mean_feedback_forward_relative_error": 0.0519376022063873, + "min_feedback_forward_cosine": 0.9957320690155029 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1615715386390686 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08872567117214203, + "mean_feedback_forward_cosine": 0.9986063913865523, + "mean_feedback_forward_relative_error": 0.049906102246181534, + "min_feedback_forward_cosine": 0.996086061000824 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16552179415225982 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08490674197673798, + "mean_feedback_forward_cosine": 0.99870783849196, + "mean_feedback_forward_relative_error": 0.04805871864611452, + "min_feedback_forward_cosine": 0.9964166879653931 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.157160287899971 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08148565143346786, + "mean_feedback_forward_cosine": 0.9988041726025668, + "mean_feedback_forward_relative_error": 0.04622622855346311, + "min_feedback_forward_cosine": 0.9967026114463806 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1610899450969696 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07781962305307388, + "mean_feedback_forward_cosine": 0.9988956841555509, + "mean_feedback_forward_relative_error": 0.04442393962632526, + "min_feedback_forward_cosine": 0.9969951510429382 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1601667630839348 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07474031299352646, + "mean_feedback_forward_cosine": 0.9989781065420671, + "mean_feedback_forward_relative_error": 0.04273093168369748, + "min_feedback_forward_cosine": 0.997226357460022 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.15819521648168564 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07225430011749268, + "mean_feedback_forward_cosine": 0.9990535009991038, + "mean_feedback_forward_relative_error": 0.04112511758099903, + "min_feedback_forward_cosine": 0.9974077343940735 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15748157766342163 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06968426704406738, + "mean_feedback_forward_cosine": 0.9991265687075528, + "mean_feedback_forward_relative_error": 0.03950692435557192, + "min_feedback_forward_cosine": 0.9975906610488892 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16175335730075835 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06721122562885284, + "mean_feedback_forward_cosine": 0.9991913286122409, + "mean_feedback_forward_relative_error": 0.03802152044055137, + "min_feedback_forward_cosine": 0.9977591037750244 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1524793500804901 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06464006751775742, + "mean_feedback_forward_cosine": 0.9992528611963446, + "mean_feedback_forward_relative_error": 0.03654412842271003, + "min_feedback_forward_cosine": 0.9979234933853149 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1604527052783966 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06187574937939644, + "mean_feedback_forward_cosine": 0.9993114037947222, + "mean_feedback_forward_relative_error": 0.03509028789333322, + "min_feedback_forward_cosine": 0.9980965852737427 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.15484731866598128 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05939973518252373, + "mean_feedback_forward_cosine": 0.999362301826477, + "mean_feedback_forward_relative_error": 0.03376985603774136, + "min_feedback_forward_cosine": 0.998246967792511 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15393349830150604 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059377092868089676, + "mean_feedback_forward_cosine": 0.9993627006357366, + "mean_feedback_forward_relative_error": 0.033757664347914135, + "min_feedback_forward_cosine": 0.9982478618621826 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08632886606931686 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05936736986041069, + "mean_feedback_forward_cosine": 0.9993629076264121, + "mean_feedback_forward_relative_error": 0.03375121339816939, + "min_feedback_forward_cosine": 0.9982483386993408 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05908495314359665 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05935855582356453, + "mean_feedback_forward_cosine": 0.9993630907752297, + "mean_feedback_forward_relative_error": 0.03374522678892721, + "min_feedback_forward_cosine": 0.9982488751411438 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.048748821417093274 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05935034528374672, + "mean_feedback_forward_cosine": 0.9993633042682301, + "mean_feedback_forward_relative_error": 0.033739580772817135, + "min_feedback_forward_cosine": 0.998249351978302 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04319611127018928 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05934198945760727, + "mean_feedback_forward_cosine": 0.9993634288961237, + "mean_feedback_forward_relative_error": 0.033734094114466144, + "min_feedback_forward_cosine": 0.998249888420105 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.038027093428373335 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05933482199907303, + "mean_feedback_forward_cosine": 0.9993636033751748, + "mean_feedback_forward_relative_error": 0.0337288758802143, + "min_feedback_forward_cosine": 0.9982502460479736 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.036271356392502785 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05932800844311714, + "mean_feedback_forward_cosine": 0.9993637550960888, + "mean_feedback_forward_relative_error": 0.03372401965951378, + "min_feedback_forward_cosine": 0.9982507228851318 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.030873922350406646 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05932128429412842, + "mean_feedback_forward_cosine": 0.9993639165704901, + "mean_feedback_forward_relative_error": 0.03371912017464638, + "min_feedback_forward_cosine": 0.9982513189315796 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.02892630367606878 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05931437388062477, + "mean_feedback_forward_cosine": 0.999364068291404, + "mean_feedback_forward_relative_error": 0.033714231103658676, + "min_feedback_forward_cosine": 0.9982515573501587 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.026974181639552116 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0593072883784771, + "mean_feedback_forward_cosine": 0.9993642265146429, + "mean_feedback_forward_relative_error": 0.03370935560458086, + "min_feedback_forward_cosine": 0.9982519745826721 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.025269123996645212 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059300776571035385, + "mean_feedback_forward_cosine": 0.9993643739006736, + "mean_feedback_forward_relative_error": 0.03370474155314944, + "min_feedback_forward_cosine": 0.9982523322105408 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.023698409345149996 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05929463356733322, + "mean_feedback_forward_cosine": 0.9993645191192627, + "mean_feedback_forward_relative_error": 0.033700264897197484, + "min_feedback_forward_cosine": 0.9982526898384094 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.021694072423279285 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059288155287504196, + "mean_feedback_forward_cosine": 0.999364686012268, + "mean_feedback_forward_relative_error": 0.03369569815695286, + "min_feedback_forward_cosine": 0.9982531070709229 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.020212705764472486 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05928143858909607, + "mean_feedback_forward_cosine": 0.9993648214773698, + "mean_feedback_forward_relative_error": 0.03369119951332157, + "min_feedback_forward_cosine": 0.9982537031173706 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.019997652606964112 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059275150299072266, + "mean_feedback_forward_cosine": 0.9993649797006087, + "mean_feedback_forward_relative_error": 0.03368665474382314, + "min_feedback_forward_cosine": 0.9982538819313049 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.018869017801880838 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059269119054079056, + "mean_feedback_forward_cosine": 0.9993651238354769, + "mean_feedback_forward_relative_error": 0.033682371489703655, + "min_feedback_forward_cosine": 0.9982541799545288 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.017170048105716706 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05926356837153435, + "mean_feedback_forward_cosine": 0.999365263635462, + "mean_feedback_forward_relative_error": 0.03367843947119333, + "min_feedback_forward_cosine": 0.9982545971870422 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.015152093956172466 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05925753712654114, + "mean_feedback_forward_cosine": 0.9993653915145181, + "mean_feedback_forward_relative_error": 0.03367414422841235, + "min_feedback_forward_cosine": 0.9982547760009766 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.015982890694737434 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05925161391496658, + "mean_feedback_forward_cosine": 0.9993655323982239, + "mean_feedback_forward_relative_error": 0.033669975645501504, + "min_feedback_forward_cosine": 0.99825519323349 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.015350015664100648 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059244535863399506, + "mean_feedback_forward_cosine": 0.9993656971237876, + "mean_feedback_forward_relative_error": 0.03366517693820325, + "min_feedback_forward_cosine": 0.9982555508613586 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.01586255150437355 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05923856422305107, + "mean_feedback_forward_cosine": 0.999365834756331, + "mean_feedback_forward_relative_error": 0.0336608559909192, + "min_feedback_forward_cosine": 0.9982557892799377 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.013963704481124878 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05923229455947876, + "mean_feedback_forward_cosine": 0.9993660081516612, + "mean_feedback_forward_relative_error": 0.033656191275539725, + "min_feedback_forward_cosine": 0.9982560873031616 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.014500608921051025 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059226661920547485, + "mean_feedback_forward_cosine": 0.9993661522865296, + "mean_feedback_forward_relative_error": 0.0336520003662868, + "min_feedback_forward_cosine": 0.9982565641403198 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.012911818793714046 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05922118201851845, + "mean_feedback_forward_cosine": 0.9993662931702354, + "mean_feedback_forward_relative_error": 0.03364786172116345, + "min_feedback_forward_cosine": 0.9982568025588989 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.012699424402415753 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05921560898423195, + "mean_feedback_forward_cosine": 0.9993664340539412, + "mean_feedback_forward_relative_error": 0.033643715862523425, + "min_feedback_forward_cosine": 0.998257040977478 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01211710981324315 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05921022593975067, + "mean_feedback_forward_cosine": 0.999366557598114, + "mean_feedback_forward_relative_error": 0.03363977646963163, + "min_feedback_forward_cosine": 0.9982572197914124 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.011623655118197202 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059204477816820145, + "mean_feedback_forward_cosine": 0.9993666941469366, + "mean_feedback_forward_relative_error": 0.0336357313462279, + "min_feedback_forward_cosine": 0.998257577419281 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.011208783966600895 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059199266135692596, + "mean_feedback_forward_cosine": 0.9993668328632008, + "mean_feedback_forward_relative_error": 0.03363182963803411, + "min_feedback_forward_cosine": 0.998258113861084 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.010452382577955723 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059193745255470276, + "mean_feedback_forward_cosine": 0.9993669683283025, + "mean_feedback_forward_relative_error": 0.03362774834544821, + "min_feedback_forward_cosine": 0.9982582926750183 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.010859078953117132 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05918765813112259, + "mean_feedback_forward_cosine": 0.9993671048771251, + "mean_feedback_forward_relative_error": 0.033623597567731686, + "min_feedback_forward_cosine": 0.9982587099075317 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009921169189661742 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059181831777095795, + "mean_feedback_forward_cosine": 0.9993672414259477, + "mean_feedback_forward_relative_error": 0.03361950904469598, + "min_feedback_forward_cosine": 0.9982592463493347 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.009995801807530224 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05917598307132721, + "mean_feedback_forward_cosine": 0.9993673833933744, + "mean_feedback_forward_relative_error": 0.033615434880961075, + "min_feedback_forward_cosine": 0.9982595443725586 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.009514382094442844 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05917006731033325, + "mean_feedback_forward_cosine": 0.999367525360801, + "mean_feedback_forward_relative_error": 0.033611282359131356, + "min_feedback_forward_cosine": 0.9982598423957825 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.009537497631013393 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05916416645050049, + "mean_feedback_forward_cosine": 0.9993676684119485, + "mean_feedback_forward_relative_error": 0.03360732892866839, + "min_feedback_forward_cosine": 0.9982601404190063 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.009508014184907078 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05915813893079758, + "mean_feedback_forward_cosine": 0.9993678255514665, + "mean_feedback_forward_relative_error": 0.03360312380404635, + "min_feedback_forward_cosine": 0.9982608556747437 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.009375937500596046 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05915252864360809, + "mean_feedback_forward_cosine": 0.9993679523468018, + "mean_feedback_forward_relative_error": 0.03359934392131188, + "min_feedback_forward_cosine": 0.9982609748840332 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008746569739431142 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059146810322999954, + "mean_feedback_forward_cosine": 0.9993680899793451, + "mean_feedback_forward_relative_error": 0.03359527747908776, + "min_feedback_forward_cosine": 0.9982613325119019 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.008365975003316999 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05914038419723511, + "mean_feedback_forward_cosine": 0.9993682514537464, + "mean_feedback_forward_relative_error": 0.03359085150401701, + "min_feedback_forward_cosine": 0.9982618093490601 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.008953102668970822 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05913488566875458, + "mean_feedback_forward_cosine": 0.9993683934211731, + "mean_feedback_forward_relative_error": 0.03358684871345759, + "min_feedback_forward_cosine": 0.9982622265815735 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.008177163695096969 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05912889167666435, + "mean_feedback_forward_cosine": 0.9993685267188332, + "mean_feedback_forward_relative_error": 0.033582697419280355, + "min_feedback_forward_cosine": 0.9982626438140869 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.008396954881548882 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05912240967154503, + "mean_feedback_forward_cosine": 0.9993686773560264, + "mean_feedback_forward_relative_error": 0.03357825619591908, + "min_feedback_forward_cosine": 0.9982631206512451 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.008055837308987976 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059114810079336166, + "mean_feedback_forward_cosine": 0.9993688615885647, + "mean_feedback_forward_relative_error": 0.03357319518585097, + "min_feedback_forward_cosine": 0.9982635974884033 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.009275788169205188 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05910937115550041, + "mean_feedback_forward_cosine": 0.9993690013885498, + "mean_feedback_forward_relative_error": 0.03356921449303627, + "min_feedback_forward_cosine": 0.998263955116272 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.007554302980676293 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0591028667986393, + "mean_feedback_forward_cosine": 0.9993691628629511, + "mean_feedback_forward_relative_error": 0.03356482491574504, + "min_feedback_forward_cosine": 0.9982640743255615 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007218919408172369 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05909779667854309, + "mean_feedback_forward_cosine": 0.9993693037466569, + "mean_feedback_forward_relative_error": 0.03356129590591246, + "min_feedback_forward_cosine": 0.998264491558075 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.006407227513724938 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05909143015742302, + "mean_feedback_forward_cosine": 0.9993694240396673, + "mean_feedback_forward_relative_error": 0.03355686375871301, + "min_feedback_forward_cosine": 0.998264729976654 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.0071185772954672575 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0590851828455925, + "mean_feedback_forward_cosine": 0.9993696093559266, + "mean_feedback_forward_relative_error": 0.03355245374312455, + "min_feedback_forward_cosine": 0.9982651472091675 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.007279688682090491 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05907873064279556, + "mean_feedback_forward_cosine": 0.9993697556582364, + "mean_feedback_forward_relative_error": 0.0335481343452226, + "min_feedback_forward_cosine": 0.9982655048370361 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006789383641332388 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05907309427857399, + "mean_feedback_forward_cosine": 0.9993698889558965, + "mean_feedback_forward_relative_error": 0.033544281125068666, + "min_feedback_forward_cosine": 0.99826580286026 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.006515305521786213 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059067174792289734, + "mean_feedback_forward_cosine": 0.9993700200861151, + "mean_feedback_forward_relative_error": 0.0335402156480334, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.006334109747409821 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906711891293526, + "mean_feedback_forward_cosine": 0.9993700222535566, + "mean_feedback_forward_relative_error": 0.0335401299494234, + "min_feedback_forward_cosine": 0.9982661008834839 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005846326352842152 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906708538532257, + "mean_feedback_forward_cosine": 0.9993700211698359, + "mean_feedback_forward_relative_error": 0.03354006827046925, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.0057507517351955174 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0590670220553875, + "mean_feedback_forward_cosine": 0.999370032007044, + "mean_feedback_forward_relative_error": 0.033540005033666435, + "min_feedback_forward_cosine": 0.9982661008834839 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.00579348510928452 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906699597835541, + "mean_feedback_forward_cosine": 0.9993700417605313, + "mean_feedback_forward_relative_error": 0.03353994405743751, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.005419869238585234 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906695872545242, + "mean_feedback_forward_cosine": 0.9993700395930897, + "mean_feedback_forward_relative_error": 0.03353988913480531, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004929324226006866 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906691029667854, + "mean_feedback_forward_cosine": 0.999370032007044, + "mean_feedback_forward_relative_error": 0.033539826871657914, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.005693616818711162 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906688794493675, + "mean_feedback_forward_cosine": 0.999370032007044, + "mean_feedback_forward_relative_error": 0.03353976836766709, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.005635865669995546 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906684696674347, + "mean_feedback_forward_cosine": 0.9993700385093689, + "mean_feedback_forward_relative_error": 0.03353971331803636, + "min_feedback_forward_cosine": 0.9982661008834839 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.004806031485870481 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066832065582275, + "mean_feedback_forward_cosine": 0.9993700417605313, + "mean_feedback_forward_relative_error": 0.03353965641422705, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.005307988227382303 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906680226325989, + "mean_feedback_forward_cosine": 0.9993700471791354, + "mean_feedback_forward_relative_error": 0.033539599027823316, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.00502156273201108 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066761285066605, + "mean_feedback_forward_cosine": 0.9993700395930897, + "mean_feedback_forward_relative_error": 0.033539544689384373, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004966159328147769 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066712856292725, + "mean_feedback_forward_cosine": 0.9993700395930897, + "mean_feedback_forward_relative_error": 0.03353949117220261, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.004718669764697552 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906669422984123, + "mean_feedback_forward_cosine": 0.9993700504302978, + "mean_feedback_forward_relative_error": 0.03353944017805836, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.004746594812879339 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066660702228546, + "mean_feedback_forward_cosine": 0.999370043927973, + "mean_feedback_forward_relative_error": 0.03353939016603611, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.004119446426257491 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906660109758377, + "mean_feedback_forward_cosine": 0.9993700591000644, + "mean_feedback_forward_relative_error": 0.03353933249176903, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0048942151668667794 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906656011939049, + "mean_feedback_forward_cosine": 0.9993700601837852, + "mean_feedback_forward_relative_error": 0.033539283995262605, + "min_feedback_forward_cosine": 0.998266339302063 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.00413149460606277 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066515415906906, + "mean_feedback_forward_cosine": 0.9993700482628562, + "mean_feedback_forward_relative_error": 0.03353923006321896, + "min_feedback_forward_cosine": 0.9982661008834839 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.004647674621641636 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0590665265917778, + "mean_feedback_forward_cosine": 0.999370049346577, + "mean_feedback_forward_relative_error": 0.03353917565704747, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.004325697136372328 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906650051474571, + "mean_feedback_forward_cosine": 0.9993700482628562, + "mean_feedback_forward_relative_error": 0.033539126661013476, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0044166714534163475 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906650051474571, + "mean_feedback_forward_cosine": 0.9993700580163436, + "mean_feedback_forward_relative_error": 0.03353908182206479, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0040626617965474725 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906646326184273, + "mean_feedback_forward_cosine": 0.9993700374256481, + "mean_feedback_forward_relative_error": 0.033539028863676566, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.004413442970365286 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906642600893974, + "mean_feedback_forward_cosine": 0.9993700536814603, + "mean_feedback_forward_relative_error": 0.033538979520513254, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0041954827751964335 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906641110777855, + "mean_feedback_forward_cosine": 0.9993700536814603, + "mean_feedback_forward_relative_error": 0.03353893079540946, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.004174803726151585 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066370129585266, + "mean_feedback_forward_cosine": 0.9993700580163436, + "mean_feedback_forward_relative_error": 0.03353887637230483, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.004461090686693788 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906631425023079, + "mean_feedback_forward_cosine": 0.9993700580163436, + "mean_feedback_forward_relative_error": 0.03353881750425154, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.004873574165888131 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066273272037506, + "mean_feedback_forward_cosine": 0.9993700591000644, + "mean_feedback_forward_relative_error": 0.033538767661560664, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.004191665274426341 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906626209616661, + "mean_feedback_forward_cosine": 0.9993700623512268, + "mean_feedback_forward_relative_error": 0.033538717124611137, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.004248902084603906 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906623601913452, + "mean_feedback_forward_cosine": 0.9993700525977395, + "mean_feedback_forward_relative_error": 0.03353866261684082, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.004457712961398065 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066202491521835, + "mean_feedback_forward_cosine": 0.9993700591000644, + "mean_feedback_forward_relative_error": 0.033538611978292464, + "min_feedback_forward_cosine": 0.998266339302063 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0043210549486428496 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906614288687706, + "mean_feedback_forward_cosine": 0.9993700623512268, + "mean_feedback_forward_relative_error": 0.03353856156834147, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0045159763356298205 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906612053513527, + "mean_feedback_forward_cosine": 0.9993700699372725, + "mean_feedback_forward_relative_error": 0.03353851396082477, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.00403653068780899 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059066079556941986, + "mean_feedback_forward_cosine": 0.9993700656023893, + "mean_feedback_forward_relative_error": 0.033538462966680525, + "min_feedback_forward_cosine": 0.998266339302063 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.004224807922951877 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906602367758751, + "mean_feedback_forward_cosine": 0.9993700677698308, + "mean_feedback_forward_relative_error": 0.03353841178627177, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.004067436493560672 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906599387526512, + "mean_feedback_forward_cosine": 0.9993700634349476, + "mean_feedback_forward_relative_error": 0.033538361875848334, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.004148853787258267 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059065982699394226, + "mean_feedback_forward_cosine": 0.9993700829419223, + "mean_feedback_forward_relative_error": 0.03353831536898559, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.004028445797674358 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906595289707184, + "mean_feedback_forward_cosine": 0.999370073188435, + "mean_feedback_forward_relative_error": 0.03353826442564076, + "min_feedback_forward_cosine": 0.9982661604881287 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.00399504036322236 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906594544649124, + "mean_feedback_forward_cosine": 0.9993700666861101, + "mean_feedback_forward_relative_error": 0.033538217800246045, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.004098718171343207 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906591936945915, + "mean_feedback_forward_cosine": 0.9993700829419223, + "mean_feedback_forward_relative_error": 0.03353816772049124, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.004304251219257712 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906587094068527, + "mean_feedback_forward_cosine": 0.9993700699372725, + "mean_feedback_forward_relative_error": 0.033538121289827603, + "min_feedback_forward_cosine": 0.9982661008834839 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.004008610866479576 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906582996249199, + "mean_feedback_forward_cosine": 0.9993700861930848, + "mean_feedback_forward_relative_error": 0.0335380723530596, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.004255639990828931 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0590657964348793, + "mean_feedback_forward_cosine": 0.9993700645186684, + "mean_feedback_forward_relative_error": 0.03353802318769422, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.004154543425813317 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906577780842781, + "mean_feedback_forward_cosine": 0.9993700721047142, + "mean_feedback_forward_relative_error": 0.0335379774343561, + "min_feedback_forward_cosine": 0.9982663989067078 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0037911088526621462 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906573683023453, + "mean_feedback_forward_cosine": 0.9993700840256431, + "mean_feedback_forward_relative_error": 0.03353792656721039, + "min_feedback_forward_cosine": 0.998266339302063 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.003857225678190589 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906572565436363, + "mean_feedback_forward_cosine": 0.9993700796907599, + "mean_feedback_forward_relative_error": 0.03353787823156877, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.003955270027332008 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906571447849274, + "mean_feedback_forward_cosine": 0.9993700710209933, + "mean_feedback_forward_relative_error": 0.03353783100504767, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.004000426731593907 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906567722558975, + "mean_feedback_forward_cosine": 0.999370078607039, + "mean_feedback_forward_relative_error": 0.033537784447385505, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.003992643709369003 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05906563252210617, + "mean_feedback_forward_cosine": 0.9993700688535517, + "mean_feedback_forward_relative_error": 0.033537734460762955, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.004213414700143039 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059065595269203186, + "mean_feedback_forward_cosine": 0.9993700775233182, + "mean_feedback_forward_relative_error": 0.0335376863791184, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0037768240635842085 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0590655654668808, + "mean_feedback_forward_cosine": 0.9993700807744806, + "mean_feedback_forward_relative_error": 0.03353764149783687, + "min_feedback_forward_cosine": 0.9982662200927734 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0033817386390268805 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059065550565719604, + "mean_feedback_forward_cosine": 0.9993700861930848, + "mean_feedback_forward_relative_error": 0.03353758930143985, + "min_feedback_forward_cosine": 0.9982662796974182 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.004372435570731759 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9212, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.4281075653076172, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3318545920, + "peak_memory_reserved_bytes": 4754243584, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.059422492980957, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 18256.374680042267, + "train_wall_s": 18250.738517045975 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250433280000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 5010580480000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json new file mode 100644 index 0000000..56e85fe --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0010051914432551712, + "innovation_negative_gradient_cosine": [ + -0.000884878623764962, + 0.0019002326298505068, + 0.0003833624068647623, + 0.0023386552929878235, + 0.0012782781850546598, + 0.0010154987685382366, + -0.0007557215867564082, + 0.0012281469535082579, + 0.003467642003670335, + 0.004312058910727501, + 0.004977978300303221, + 0.0059320577420294285, + 0.0031003656331449747, + 0.00814433116465807, + 0.012635820545256138, + 0.01465326827019453, + 0.016317065805196762, + 0.021564975380897522, + 0.028595589101314545 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.000884878623764962, + 0.0019002326298505068, + 0.0003833624068647623, + 0.0023386552929878235, + 0.0012782781850546598, + 0.0010154987685382366, + -0.0007557215867564082, + 0.0012281469535082579, + 0.003467642003670335, + 0.004312058910727501, + 0.004977978300303221, + 0.0059320577420294285, + 0.0031003656331449747, + 0.00814433116465807, + 0.012635820545256138, + 0.01465326827019453, + 0.016317065805196762, + 0.021564975380897522, + 0.028595589101314545 + ], + "teaching_negative_gradient_cosine": [ + -0.000884878623764962, + 0.0019002326298505068, + 0.0003833624068647623, + 0.0023386552929878235, + 0.0012782781850546598, + 0.0010154987685382366, + -0.0007557215867564082, + 0.0012281469535082579, + 0.003467642003670335, + 0.004312058910727501, + 0.004977978300303221, + 0.0059320577420294285, + 0.0031003656331449747, + 0.00814433116465807, + 0.012635820545256138, + 0.01465326827019453, + 0.016317065805196762, + 0.021564975380897522, + 0.028595589101314545 + ], + "wall_s": 0.10386037826538086 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.5232079847717284 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 3.086358903579712 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 5.356443600463868 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.332649574584961 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 11.553366081848145 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 15.536432564086914 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 20.713728800048827 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 26.808758018798827 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 33.451930782470704 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 43.32454557983399 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 47.37369536376953 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 58.366754450683594 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 63.66044085693359 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 80.79366400146485 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 89.35128830627441 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 104.20707355102539 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 112.15316399902343 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 131.0725962915039 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 146.9820818774414 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 162.69177174804688 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 189.16082287597655 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 200.92569576416017 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 238.4890877685547 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 236.53758495117188 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 274.56765765625 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 315.4286769433594 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 303.08119244140624 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 338.9314815234375 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 372.47922338867187 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 393.44156157226564 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 439.79392638671874 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 426.92012123046874 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 519.1147138671874 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 543.4393654296875 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 566.3028992773437 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 614.261176328125 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 711.8948611132812 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 732.9045906933594 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 732.0083269140625 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 785.885091015625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 848.0388903515625 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 932.2692114453125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 892.872412109375 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 961.143112421875 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 1070.3008855859375 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 1050.7108417578124 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 1028.1947622265625 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 1054.1042734375 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 1219.4650396875 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1216.48408203125 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1321.9611243554687 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1381.535570703125 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1497.6194152539063 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1503.97397734375 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1657.0008198828125 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1588.29275328125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1688.0786598828124 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 1732.75021234375 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1957.915803359375 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 1894.246257421875 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 1994.7383422265625 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 2077.9824059375 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 2072.7607075 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 2235.696065078125 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 2297.305461328125 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2229.95345859375 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2609.048402890625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 2470.129161015625 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 2576.2696977734377 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2400.776232109375 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 2780.13626671875 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 2902.25096015625 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 3203.1363859375 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 2849.149858671875 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 3157.0732768359376 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 3252.20273203125 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 3054.77422296875 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 3548.091424453125 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 3741.11722765625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 3364.16918453125 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 3776.60609140625 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3695.4534409375 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3781.0818490625 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 4011.38095234375 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 4079.54671640625 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 4359.65175375 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 4350.933033125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 4334.60786390625 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 4892.78349609375 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 4625.3204778125 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 4578.29714671875 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 5343.80255359375 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 4621.24079125 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 4941.87048484375 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 5435.19049 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 4987.83706609375 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 5581.69119203125 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 5146.3901084375 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 5657.07179296875 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 6263.5825775 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1294.296304609375 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 906.5465127929688 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 794.4426366015625 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 731.5252470117188 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 688.6400637304688 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 643.1145382421875 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 607.4473750585937 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 580.257162890625 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 539.6933870214843 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 518.1951058203125 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 491.1930346777344 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 480.9726749023437 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 439.992346640625 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 421.9247789453125 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 438.31238682617186 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 448.0504717480469 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 443.4344926171875 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 420.33937994140626 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 435.94316953125 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 425.4205125976562 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 429.7772877441406 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 426.15123880859375 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 424.79062931640624 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 405.5273923828125 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 414.800235703125 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 409.8256136621094 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 427.84816609375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 420.3188088867187 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 449.41646658203126 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 430.15067587890627 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 448.34247743164065 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 404.2150330859375 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 445.31830255859376 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 420.33240436523437 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 464.25871956054687 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 451.64062458984375 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 400.7794763867187 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 443.15128728515623 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 464.11758080078124 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 435.2545360546875 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 476.9718269140625 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 408.586395078125 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 450.2365105859375 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 501.0252286621094 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 468.44268588867186 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 442.7631327148438 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 467.3138282128906 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 473.6360938574219 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 501.6530146435547 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 449.1373256542969 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 160.08695232910156 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 140.47828806152344 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 131.62066892089842 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 122.39664599121093 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 117.20858522949219 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 111.00679817871094 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 107.39616866210937 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 101.54270803222656 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 99.27490577392578 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 94.62851961669922 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 91.61981069091797 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 88.54825022460938 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 86.99606296386719 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 85.45751216064453 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 81.17958895996094 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 79.90534064208984 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 77.40922764404297 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 77.25150076904296 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 73.21587929443359 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 72.43009603759765 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 72.41567712158204 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 69.02928202392579 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 67.33243247558593 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 68.03048149902344 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 66.97131014648437 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 64.10073219726563 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 67.55217951171875 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 65.45841945800781 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 64.84410902587891 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 65.5598119555664 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 62.593642211914066 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 62.848734842529296 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 63.53649013183594 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 60.65613086791992 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 62.11923370117187 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 62.43092033935547 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 61.30139270141601 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 59.336867834472656 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 60.86013489013672 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 61.31213500488281 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 56.716611485595706 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 62.33711942626953 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 61.05544593017578 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 57.704346300048826 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 55.51626229858398 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 63.45401997802735 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 56.87155424072266 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 58.41696393310547 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 57.84488670166016 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 55.743933559570316 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.372, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 52.914790234375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7046947479248047, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3656.697639465332, + "train_wall_s": 3655.9695632457733 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json new file mode 100644 index 0000000..398e9b6 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0003854467940982431, + "innovation_negative_gradient_cosine": [ + -0.0010154233314096928, + 0.0011164090828970075, + 0.0018438097322359681, + 0.001890377956442535, + -0.0010229444596916437, + -0.0004995482158847153, + 0.0019841650500893593, + 0.0077583095990121365, + 0.0021534061525017023, + 0.0015168748795986176, + 0.005302063189446926, + 0.0065992907620966434, + 0.004412676207721233, + 0.02175547368824482, + 0.022844579070806503, + 0.022070160135626793, + 0.029950225725769997, + 0.02953408658504486, + 0.039382897317409515 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0010154233314096928, + 0.0011164090828970075, + 0.0018438097322359681, + 0.001890377956442535, + -0.0010229444596916437, + -0.0004995482158847153, + 0.0019841650500893593, + 0.0077583095990121365, + 0.0021534061525017023, + 0.0015168748795986176, + 0.005302063189446926, + 0.0065992907620966434, + 0.004412676207721233, + 0.02175547368824482, + 0.022844579070806503, + 0.022070160135626793, + 0.029950225725769997, + 0.02953408658504486, + 0.039382897317409515 + ], + "teaching_negative_gradient_cosine": [ + -0.0010154233314096928, + 0.0011164090828970075, + 0.0018438097322359681, + 0.001890377956442535, + -0.0010229444596916437, + -0.0004995482158847153, + 0.0019841650500893593, + 0.0077583095990121365, + 0.0021534061525017023, + 0.0015168748795986176, + 0.005302063189446926, + 0.0065992907620966434, + 0.004412676207721233, + 0.02175547368824482, + 0.022844579070806503, + 0.022070160135626793, + 0.029950225725769997, + 0.02953408658504486, + 0.039382897317409515 + ], + "wall_s": 0.10302209854125977 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.5068225940704347 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 3.515857922210693 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 5.684407122192383 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.086542698059082 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 12.036429146728516 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 16.34205907470703 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 21.01639781677246 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 27.02610977294922 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 35.431440301513675 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 41.92359442016602 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 50.71551621459961 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 59.637366508789064 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 74.75878812622071 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 85.2104129663086 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 89.33222989990234 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 114.89599761596679 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 122.97920129638672 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 137.3023895336914 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 163.0888639453125 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 180.36466465332032 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 204.6404056591797 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 218.40419555664062 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 243.00024526367187 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 254.7752802685547 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 292.07824676757815 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 329.8202194238281 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 339.64070388671877 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 399.37932737304686 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 428.4542909765625 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 410.7046250390625 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 491.46051373046873 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 485.348603984375 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 511.99018515625 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 570.5548893359374 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 599.0122534863282 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 638.6346241601562 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 655.8679623828125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 732.1800444726563 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 793.89914203125 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 770.4782406445313 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 838.25508484375 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 868.5379801171875 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 977.8729441796875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 943.8888183007813 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 1085.8748257226562 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 1121.91362421875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 1149.345110078125 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 1268.7161190429688 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 1291.9986146484375 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1284.3497418359375 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1399.3627420703126 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1612.1165806640624 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1648.3374513671874 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1674.036724765625 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1560.2893973828125 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1711.9624044921875 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1712.132676171875 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 2055.0074187890623 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1829.655229375 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 2050.5391868359375 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 2019.0894387890626 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 1942.0169362109375 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 2274.977286015625 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 2283.78664203125 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 2264.264823359375 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2488.3162587109373 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2646.7587425 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 2517.2651884375 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 2730.358718671875 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2989.476571875 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 2719.420979453125 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 2725.74516234375 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 3095.27936921875 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 3115.5268525 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 3076.67187625 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 2864.8534446875 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 3407.74620171875 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 3591.5484866796874 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 3505.355758515625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 3667.21988140625 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 3818.790492109375 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3990.26935640625 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3560.0794140625 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 4143.06030046875 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 4049.86179421875 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 4096.51183484375 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 4365.442724296875 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 4307.64304 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 4608.90662203125 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 4627.5656590625 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 5346.3826034375 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 4928.42728703125 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 4834.4236396875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 4986.335101875 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 5177.383709609375 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 5380.46290078125 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 5570.4661796875 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 5715.67653625 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 5869.27433859375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 5563.03640296875 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1193.3901241796875 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 901.1035458203125 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 804.7639102734375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 716.0620738085937 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 647.3206634179687 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 600.5981295703125 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 555.0965403808593 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 531.3540760351563 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 524.3530202246094 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 471.2825883300781 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 452.90929244140625 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 438.8323709179688 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 423.5165996386719 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 426.702331640625 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 423.29922956054685 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 404.40631328125 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 428.2123980078125 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 400.1490480859375 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 434.12222455078125 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 412.13692862304686 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 424.259929921875 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 459.1871238867187 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 402.66018516601565 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 438.0382576464844 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 437.2812224804687 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 451.2520609863281 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 421.0700023046875 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 458.96626716796874 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 412.06519220703126 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 474.48770896484376 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 411.5285243847656 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 475.40324076171873 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 433.22330268066406 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 463.81109161132815 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 449.84764244140626 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 425.2446093261719 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 507.0734819042969 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 400.82680666015625 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 497.38916314453127 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 468.95010198242187 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 465.1990760351562 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 456.9041644140625 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 470.43679358398435 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 462.1703258789062 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 444.6902553173828 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 486.78014923828124 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 454.7092383691406 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 451.41220109375 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 450.31422676757813 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 467.62194791992187 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 140.5188960180664 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 122.81540368164063 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 116.09231616699219 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 107.69791808105468 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 102.92118276855469 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 97.91107972167968 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 94.57610254394531 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 91.08143123291016 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 87.3799490209961 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 83.78194811523437 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 81.5281500390625 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 76.96467525634766 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 74.48964189941407 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 74.4926614038086 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 72.4107512121582 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 69.0355158581543 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 69.6121717578125 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 66.27636252441407 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 66.68412861816407 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 66.87026070922852 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 63.41542741210937 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 62.05930561645508 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 60.717706286621095 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 60.02907170166016 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 59.69362048095703 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 62.324915122070315 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 58.233677922363285 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 55.237775908203126 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 57.21391374755859 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 59.198869759521486 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 55.32839068359375 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 54.57799655761719 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 57.88844884033203 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 56.9704599609375 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 53.511298244628904 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 52.950861934814455 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 54.97057058349609 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 57.945177227783205 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 52.78738121826172 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 53.15324090576172 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 52.12365180419922 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 52.32365452270508 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 50.616903367919925 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 53.57607133056641 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 53.6800094543457 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 51.921997236328124 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 53.47778112060547 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 51.79992364501953 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 54.57040306396485 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 51.30855579345703 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3374, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 37.194462109375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7038471698760986, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3643.292373418808, + "train_wall_s": 3642.5641095638275 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json new file mode 100644 index 0000000..baee899 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0008285329871190091, + "innovation_negative_gradient_cosine": [ + -0.0009353021159768105, + 0.0011565475724637508, + 0.000651370850391686, + 0.0012092051329091191, + 0.0015328937442973256, + 0.0013564827386289835, + 6.672111339867115e-06, + 0.004789423197507858, + 0.0019560717046260834, + 0.003033566987141967, + 0.004105382598936558, + 0.0018228725530207157, + 0.00795711949467659, + 0.016370896250009537, + 0.020950693637132645, + 0.020420797169208527, + 0.025357890874147415, + 0.02774948999285698, + 0.045218706130981445 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0009353021159768105, + 0.0011565475724637508, + 0.000651370850391686, + 0.0012092051329091191, + 0.0015328937442973256, + 0.0013564827386289835, + 6.672111339867115e-06, + 0.004789423197507858, + 0.0019560717046260834, + 0.003033566987141967, + 0.004105382598936558, + 0.0018228725530207157, + 0.00795711949467659, + 0.016370896250009537, + 0.020950693637132645, + 0.020420797169208527, + 0.025357890874147415, + 0.02774948999285698, + 0.045218706130981445 + ], + "teaching_negative_gradient_cosine": [ + -0.0009353021159768105, + 0.0011565475724637508, + 0.000651370850391686, + 0.0012092051329091191, + 0.0015328937442973256, + 0.0013564827386289835, + 6.672111339867115e-06, + 0.004789423197507858, + 0.0019560717046260834, + 0.003033566987141967, + 0.004105382598936558, + 0.0018228725530207157, + 0.00795711949467659, + 0.016370896250009537, + 0.020950693637132645, + 0.020420797169208527, + 0.025357890874147415, + 0.02774948999285698, + 0.045218706130981445 + ], + "wall_s": 0.12566781044006348 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.6916630249786375 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 3.1234489201354982 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 5.98929384399414 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.105805624694824 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 11.538696815185547 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 16.250854111633302 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 19.46256600463867 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 26.818485535888673 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 32.77349362548828 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 39.53968194580078 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 45.77936697265625 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 58.16394977539063 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 66.18473546386718 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 78.83503866699219 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 86.3635105078125 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 95.0407410546875 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 107.21239222900391 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 128.04616498535157 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 149.76198973632813 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 172.02814325683593 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 190.65955872558592 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 188.47470913085937 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 200.3554741015625 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 220.64422837890626 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 237.45402358398437 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 256.9474181494141 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 277.12291538085935 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 295.2959725878906 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 311.341398203125 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 333.3710159960938 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 382.4080616015625 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 396.0423643261719 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 390.58979763671874 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 436.208059765625 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 480.45594244140625 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 541.5972395117187 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 528.4664376171875 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 583.1408534960938 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 614.4528769921875 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 664.7279025390625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 732.3504648242188 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 701.6167083203125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 851.5649462304688 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 815.2471684375 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 877.268365078125 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 930.6870291015625 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 983.5310730859375 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 979.5800359375 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 934.2593817578125 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1079.95480984375 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1052.201548671875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1147.88641375 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1149.809829609375 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1293.8795182421875 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1307.43599734375 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1306.2825561328125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1400.6581353125 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 1495.918426953125 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1634.0218153125 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 1572.8116048828124 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 1477.91943578125 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 1751.65383234375 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 1663.04731765625 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 1988.0621076171874 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 1764.8227140234376 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2160.3468409765624 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2041.295907109375 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 1778.82399515625 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 2092.801682734375 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2352.102005 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 2183.911658984375 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 2306.172326796875 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 2314.36352390625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 2558.8601367578126 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 2422.2948581640626 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 2448.32666484375 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 2667.2972108984377 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 2643.30885875 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 2935.8231928125 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 2792.73163734375 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 2801.67680078125 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3121.9187403125 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3058.5406184375 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 2982.79893328125 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 3130.38434546875 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 3178.91528828125 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 3222.1684484375 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 3380.3048425 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 3553.93340265625 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 3898.7508175 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 3509.245996875 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 3729.505656328125 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 3738.689292109375 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 4285.6902996875 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 3892.4513234765627 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 4566.9196865625 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 4484.4360309375 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 4162.1949146875 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 4350.78345109375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 4286.676404375 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1034.5034601171874 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 691.3320501953125 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 589.5045511328125 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 535.446396328125 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 500.77169568359375 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 474.63076671875 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 451.45938520507815 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 434.28193009765624 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 400.94140069335936 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 388.41041271484374 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 373.98031665039065 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 370.25313810546874 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 367.33843630859377 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 353.6634990234375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 342.0300949902344 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 346.350079296875 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 369.759463359375 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 304.771866640625 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 357.7160865527344 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 336.3204292480469 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 319.9236466894531 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 364.6893849902344 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 339.5355904492188 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 334.9873714160156 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 345.6425043652344 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 339.8361540576172 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 367.39203767578124 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 344.979698203125 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 334.61052381835935 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 366.74309252929686 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 325.3127012402344 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 340.94689490234373 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 338.9773078125 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 325.57656227539064 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 321.4540401464844 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 343.552309140625 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 333.6868705029297 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 371.4429591015625 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 348.3975844189453 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 331.80154568359376 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 346.2748961816406 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 340.55735438476563 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 387.8134982421875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 326.7488217089844 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 366.3015127246094 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 385.80695004394533 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 355.3764542285156 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 372.8823209765625 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 332.7078068359375 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 360.0113352734375 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 131.34461179199218 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 108.80747685058594 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 100.66301383300781 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 94.52819634521484 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 87.97953765869141 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 85.77151513916016 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 82.03563292114258 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 78.74349596191406 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 76.2903619970703 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 73.58587282348633 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 71.40343949707031 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 69.75140670043945 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 67.43331412109374 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 66.4655532397461 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 64.4736221081543 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 62.45239664550781 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 60.40092920654297 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 60.63630566894531 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 57.50094825683594 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 57.094085632324216 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 55.83090142944336 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 56.35826234619141 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 55.12465389892578 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 54.59146714355469 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 52.23197322265625 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 53.748194729003906 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 53.08256021972656 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 54.18801360717774 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 50.570823845214846 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 51.9733895690918 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 53.15994084472656 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 53.409358203125 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 51.330983424072265 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 51.64404582275391 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 50.77545739501953 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 52.275024084472655 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 50.963492211914065 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 49.07951643554688 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 48.67352672363281 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 49.29971404785156 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 50.59038954223633 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 50.11300712158203 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 49.198302316894534 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 47.84505870117187 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 46.51261842895508 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 48.02750996948242 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 48.127764237060546 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 49.32261536254883 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 47.53660641601562 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 47.61768265869141 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.2628, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 56.3721171875, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7035977840423584, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3643.039246082306, + "train_wall_s": 3642.31250166893 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json new file mode 100644 index 0000000..3442c49 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.001026772321589912, + "innovation_negative_gradient_cosine": [ + 0.00043954188004136086, + 0.0005074266227893531, + 0.0011681623291224241, + 0.0016458929749205709, + 0.0016051034908741713, + 0.0007945066317915916, + 0.001554696587845683, + 0.006644959561526775, + 0.005999387241899967, + 0.0019136364571750164, + 0.002381813945248723, + 0.004437652416527271, + 0.004399558529257774, + 0.01699688658118248, + 0.01732775568962097, + 0.018447382375597954, + 0.020006468519568443, + 0.02677552029490471, + 0.027215570211410522 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.00043954188004136086, + 0.0005074266227893531, + 0.0011681623291224241, + 0.0016458929749205709, + 0.0016051034908741713, + 0.0007945066317915916, + 0.001554696587845683, + 0.006644959561526775, + 0.005999387241899967, + 0.0019136364571750164, + 0.002381813945248723, + 0.004437652416527271, + 0.004399558529257774, + 0.01699688658118248, + 0.01732775568962097, + 0.018447382375597954, + 0.020006468519568443, + 0.02677552029490471, + 0.027215570211410522 + ], + "teaching_negative_gradient_cosine": [ + 0.00043954188004136086, + 0.0005074266227893531, + 0.0011681623291224241, + 0.0016458929749205709, + 0.0016051034908741713, + 0.0007945066317915916, + 0.001554696587845683, + 0.006644959561526775, + 0.005999387241899967, + 0.0019136364571750164, + 0.002381813945248723, + 0.004437652416527271, + 0.004399558529257774, + 0.01699688658118248, + 0.01732775568962097, + 0.018447382375597954, + 0.020006468519568443, + 0.02677552029490471, + 0.027215570211410522 + ], + "wall_s": 0.10304927825927734 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.597743264465332 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 3.368424224243164 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 5.0752776852417 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.323098348083496 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 12.31845982788086 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 16.29355094329834 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 20.603095179443358 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 29.07681758544922 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 35.548073359375 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 45.14130805419922 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 57.08424531738281 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 68.2577324243164 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 79.36335182373047 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 99.69380631713867 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 109.6113326611328 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 126.63547584472656 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 155.73404821777345 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 160.2812972314453 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 175.9889497216797 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 198.99891857421875 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 227.77185265625 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 243.32541502929686 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 267.96958142089846 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 293.82499244140627 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 295.66653302246095 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 319.86489938964843 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 379.9896616796875 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 384.0649412890625 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 436.53419962890626 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 454.373347890625 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 471.4458005566406 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 510.7886199609375 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 573.48722703125 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 589.5229011621094 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 613.3814142773438 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 639.5028542578125 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 722.1255887304687 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 750.0486928515625 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 821.027588984375 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 854.4821410546875 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 846.2923630664062 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 944.7046792578125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 1039.3222859570312 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 1048.7743979882812 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 1103.526996640625 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 1178.75220375 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 1156.3368054492187 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 1307.992951171875 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 1206.647822265625 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1420.3804480078124 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1349.799582890625 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1552.5224446679688 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1570.208374296875 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1545.1350866796874 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1732.9275464453126 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1639.574808828125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1762.685194921875 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 1691.6347921484376 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1890.61962625 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 1954.956363125 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 2086.688307734375 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 2327.78829984375 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 2232.0103207421876 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 2201.0438309375 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 2411.561822890625 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2240.848030546875 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2465.246695625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 2540.17220703125 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 2630.3325715625 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2614.82955703125 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 2790.60815625 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 2712.09315296875 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 2759.56806484375 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 2951.730076953125 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 3185.164907734375 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 3156.181993828125 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 3298.50851984375 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 3786.514894375 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 3540.58271890625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 3837.29801390625 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 3901.723029296875 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3993.34547984375 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3677.507210078125 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 4320.35942125 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 3959.435968515625 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 4117.64078765625 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 3946.00810875 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 4256.0377534375 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 4551.68098046875 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 4649.0053834375 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 4658.92454953125 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 5086.8264853125 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 4588.0969928125 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 5325.870966875 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 4527.5314653125 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 5515.834625625 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 5089.352583828125 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 5323.62668078125 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 5106.75947984375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 5020.8132815625 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1185.758274765625 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 864.0628663671876 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 748.6196676953125 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 671.3818684179688 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 616.1261779296875 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 568.0862041015625 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 536.3782029296875 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 509.1223760742188 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 476.9261923535156 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 455.9505439746094 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 471.44886276367185 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 452.33764107421877 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 437.6467414257813 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 407.5566308984375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 422.86549296875 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 415.95467072265626 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 423.545407265625 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 385.2936622949219 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 429.5061202050781 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 378.22622220703124 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 412.6712507226562 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 407.872823671875 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 425.10266803710937 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 416.83781041015624 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 403.8425468847656 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 403.4748011816406 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 395.99779684570314 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 454.81411276367186 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 390.7997586035156 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 460.4342845703125 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 430.23412443359376 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 447.54191185546875 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 419.99701744140623 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 442.2202359375 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 399.64367326171873 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 469.29694830078125 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 406.5749193457031 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 409.1965287402344 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 453.96415654296874 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 425.6436191796875 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 438.41210352539065 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 416.19182708984374 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 464.909064921875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 408.803461953125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 420.28573818359376 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 481.28631602539065 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 436.3826825 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 468.26839637695315 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 441.45955795898436 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 474.4885379296875 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 150.8306891503906 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 126.03814733886719 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 116.59625692871094 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 110.63184271728515 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 104.42011940917969 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 98.08100389648438 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 94.29755151611329 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 91.38015209472657 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 87.2267155871582 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 86.04973744873047 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 83.78250428955079 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 80.57094376708984 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 76.09295771484375 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 73.65551808837891 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 72.52402937988282 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 71.30078241943359 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 68.89115245117188 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 66.90169017822265 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 66.67080534667969 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 65.27129294677735 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 64.8155755065918 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 62.95781930908203 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 61.859475665283206 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 62.85311723510742 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 58.535931169433596 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 59.294098100585934 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 59.13398171020508 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 62.26549455566406 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 59.120550153808594 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 57.00106326049805 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 58.11168995361328 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 55.04777943725586 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 55.54710742675781 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 57.45380103027344 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 57.00285368164062 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 52.12349615722656 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 57.2096112866211 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 52.982457048339846 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 59.01710743896484 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 52.46784541748047 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 55.75999948364258 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 53.59568301513672 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 51.69608590820312 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 54.38446314697266 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 52.21307955200195 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 53.86592844116211 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 55.33897862182617 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 51.4826995703125 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 50.6251251171875 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 51.178080053710936 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3417, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 44.02001796875, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7042059898376465, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3654.315833568573, + "train_wall_s": 3653.588398694992 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json new file mode 100644 index 0000000..f0c00a1 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.000274311537699153, + "innovation_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "teaching_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "wall_s": 0.12834787368774414 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.430827777557373 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 2.8695908666992187 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 4.850728899841308 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.524278156127929 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 11.590560819702148 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 15.759647039794922 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 22.13480206237793 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 26.629495576171873 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 36.430580815429686 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 44.20328767089844 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 48.577757346191405 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 56.446661684570316 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 68.82190098266602 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 73.90951887939453 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 90.26532935302734 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 111.07496459472657 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 114.86494052490234 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 142.34349526855468 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 168.44054090087891 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 179.05319467773438 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 202.31314580078126 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 205.62455334960939 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 248.4839746875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 247.82219323242188 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 279.3399982861328 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 306.9608862109375 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 346.06842751953127 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 376.8380232617188 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 432.8407747265625 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 457.6128712109375 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 466.56698500976563 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 519.24828078125 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 571.6328147070312 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 572.9210324511719 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 602.8327405859375 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 670.990899296875 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 746.16743203125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 728.067555703125 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 754.1594594726563 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 925.0366741796875 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 838.8716380078125 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 942.319340859375 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 980.6203349609375 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 1030.9713324414063 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 1134.9342415234375 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 1201.5530361328124 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 1126.673959609375 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 1320.0336044921876 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 1355.6681817578126 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1343.8974589453126 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1293.3335163671875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1492.314264765625 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1695.872830703125 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1602.5963324609374 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1713.265025546875 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1812.542683203125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1860.6483206445312 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 1952.0459861328125 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1966.81673078125 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 2140.2526516796875 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 2067.9922076171874 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 2276.8285640234376 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 2251.3169115625 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 2529.34507828125 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 2340.8230924609375 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2649.29321765625 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2679.8316846875 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 2818.46077625 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 3011.803917421875 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2959.873146484375 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 3122.5627303125 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 3118.783563828125 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 3193.22344640625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 3433.978434609375 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 3412.4790053125 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 3107.52170140625 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 3805.391980703125 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 3394.3623228515626 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 3846.74855015625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 3939.635757890625 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 3881.772237421875 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3870.17540359375 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3981.95599703125 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 4510.165597421875 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 4949.16415359375 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 4467.54481765625 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 4262.94804125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 4779.1370278125 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 4564.3299187890625 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 4644.84135625 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 4634.471859375 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 5272.44126296875 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 5398.831680546875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 5311.88040921875 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 5491.19182453125 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 5483.284246875 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 5769.0386840625 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 5507.14149140625 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 5734.120633125 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 6269.34827296875 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1487.558666796875 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 950.5722432617188 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 832.749403671875 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 753.5799643359375 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 705.256015078125 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 674.2632983984375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 637.6445683203125 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 599.0539654101563 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 575.545034296875 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 534.5734385546875 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 547.174887890625 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 502.4583783984375 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 489.33109234375 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 460.824575859375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 476.13343795898436 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 473.103417109375 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 468.42028130859376 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 451.5446798828125 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 448.97799533203124 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 436.08307154296875 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 452.778623984375 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 465.5214141796875 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 458.1043487207031 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 448.01370393554686 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 455.8718875292969 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 441.31153591796874 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 471.42664501953124 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 462.54126444335935 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 478.6115515136719 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 469.2341409375 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 454.92755802734376 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 441.7504991113281 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 506.769214453125 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 431.8849821191406 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 464.18567890625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 453.88623982421876 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 453.206499375 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 476.1337943457031 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 530.7403985253907 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 500.3454430273438 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 441.26398439453124 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 483.9563454199219 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 467.760279453125 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 493.96308111328125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 475.72507780273435 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 453.8898660546875 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 535.6345719726562 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 478.78393487304686 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 532.946368515625 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 512.34838421875 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 179.30683587402345 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 152.58808825195314 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 140.42047415527344 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 130.18362895507812 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 123.7321546875 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 117.87116739746094 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 113.00177812744141 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 107.92502445800781 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 103.53310265136719 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 100.67651490722656 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 96.55011247802734 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 92.68933987060547 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 90.2942857080078 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 86.75207002685546 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 85.0101225024414 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 82.25027728271485 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 80.04829027099609 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 79.29926852783203 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 75.78352020751953 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 73.82204443847657 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 72.57129687255859 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 73.18166558715821 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 68.87338528564453 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 68.01583039794922 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 67.75869918701171 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 66.2078173413086 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 66.47758010498048 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 63.72786291503906 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 66.09409375976563 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 67.07359419921875 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 63.96962786376953 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 62.141143621826174 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 62.4706698059082 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 60.51575397949219 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 62.271037517089844 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 61.85910586181641 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 62.088958939208986 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 61.23988405395508 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 57.6482016003418 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 60.53348528076172 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 57.87255747314453 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 59.5084599609375 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 57.39036342163086 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 59.12267486328125 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 57.19819190673828 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 56.51260213256836 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 57.803850379638675 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 57.83694255371094 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 54.8499035168457 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 55.30137119018555 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.28, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 74.0652265625, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7102277278900146, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3641.387490272522, + "train_wall_s": 3640.65451335907 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json b/results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json new file mode 100644 index 0000000..cb2b886 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json @@ -0,0 +1,2089 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 3.074540582019836e-05, + "innovation_negative_gradient_cosine": [ + -0.00036826980067417026, + 2.6044523110613227e-05, + 0.0008353812154382467, + -0.0008176610572263598, + 0.0013316046679392457, + -0.0018267796840518713, + -0.00035030735307373106, + 0.00024559281882829964, + -0.0017333189025521278, + 0.0029651676304638386, + -0.0007047887775115669, + 0.0012012019287794828, + -0.0003103498602285981, + 0.0016303639858961105, + 0.0007953367312438786, + 0.002465467667207122, + 0.0007311954977922142, + -0.003911031410098076, + 0.0066788699477910995, + 0.0017323954962193966, + 0.001733468729071319, + 0.007426526863127947, + 0.009960150346159935, + 0.009964510798454285, + 0.013116281479597092, + 0.00614601606503129, + 0.013978250324726105, + 0.008247887715697289, + 0.015481583774089813, + 0.013867078348994255, + 0.01987851783633232 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.00036826980067417026, + 2.6044523110613227e-05, + 0.0008353812154382467, + -0.0008176610572263598, + 0.0013316046679392457, + -0.0018267796840518713, + -0.00035030735307373106, + 0.00024559281882829964, + -0.0017333189025521278, + 0.0029651676304638386, + -0.0007047887775115669, + 0.0012012019287794828, + -0.0003103498602285981, + 0.0016303639858961105, + 0.0007953367312438786, + 0.002465467667207122, + 0.0007311954977922142, + -0.003911031410098076, + 0.0066788699477910995, + 0.0017323954962193966, + 0.001733468729071319, + 0.007426526863127947, + 0.009960150346159935, + 0.009964510798454285, + 0.013116281479597092, + 0.00614601606503129, + 0.013978250324726105, + 0.008247887715697289, + 0.015481583774089813, + 0.013867078348994255, + 0.01987851783633232 + ], + "teaching_negative_gradient_cosine": [ + -0.00036826980067417026, + 2.6044523110613227e-05, + 0.0008353812154382467, + -0.0008176610572263598, + 0.0013316046679392457, + -0.0018267796840518713, + -0.00035030735307373106, + 0.00024559281882829964, + -0.0017333189025521278, + 0.0029651676304638386, + -0.0007047887775115669, + 0.0012012019287794828, + -0.0003103498602285981, + 0.0016303639858961105, + 0.0007953367312438786, + 0.002465467667207122, + 0.0007311954977922142, + -0.003911031410098076, + 0.0066788699477910995, + 0.0017323954962193966, + 0.001733468729071319, + 0.007426526863127947, + 0.009960150346159935, + 0.009964510798454285, + 0.013116281479597092, + 0.00614601606503129, + 0.013978250324726105, + 0.008247887715697289, + 0.015481583774089813, + 0.013867078348994255, + 0.01987851783633232 + ], + "wall_s": 0.14246678352355957 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 6.293598889007568 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 7.412033297271728 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 10.737898612060548 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 16.133542594604492 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 22.19608327026367 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 30.915053834228516 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 39.64462054199219 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 57.555933121337894 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 74.06779753051758 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 94.07128509277344 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 111.10326148925782 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 141.19566643066406 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 170.659443046875 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 194.02587569335938 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 233.4889348339844 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 271.0032402294922 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 314.8979043945312 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 347.94102638671876 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 409.38084181640625 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 441.446261484375 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 516.6539449804687 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 539.8580530175781 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 580.1104050195313 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 676.0906620703125 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 765.2543776171875 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 865.8800613671875 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 933.8132362890625 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 1060.9912740625 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 978.6808122265625 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 1180.31563515625 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 1244.2332968359376 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 1351.088999765625 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 1416.4225203125 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 1457.549191484375 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 1664.370351171875 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 1552.4956359375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 1821.0322580078125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 1990.8570221875 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 1969.9385286328124 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 2325.989310234375 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 2438.87965421875 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 2770.6505803125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 2711.5188660546874 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 2650.04287421875 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 2918.512974765625 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 3005.5448178125 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 3492.73231265625 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 3662.0783178125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 3463.054914609375 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 3419.4710433984374 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 3948.77810875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 4149.43587421875 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 4289.252799921875 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 4122.716543242187 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 4666.4023015625 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 4463.53142515625 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 4326.788139375 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 5110.3531796875 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 5457.86820171875 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 5612.76991125 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 5747.9700634375 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 5802.20757921875 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 5967.9336390625 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 6246.990884921875 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 6950.7405046875 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 6945.5849509375 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 6585.1524090625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 6822.074729375 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 7025.1091890625 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 8168.0862190625 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 8214.2977309375 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 7491.4265615625 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 8174.256045625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 8563.91816875 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 8911.6764040625 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 8645.0338546875 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 8843.74548875 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 9447.9077440625 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 9591.966471875 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 9565.742981875 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 10660.2193578125 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 11309.25382125 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 10361.5855065625 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 10960.4349828125 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 11115.4821184375 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 11953.2073675 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 11733.153768125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 11426.63758875 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 13487.809304375 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 13246.3208703125 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 12576.546934375 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 14237.348911875 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 13395.600990625 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 15017.1467109375 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 14967.916365625 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 14193.336595625 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 13713.72429125 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 16357.4106659375 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 14531.0011725 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 15943.223449375 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 3344.21362453125 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 2304.325952890625 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 2105.208774375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 2003.5612227734375 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 1844.6388519140626 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 1716.0516268359374 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 1653.430671640625 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 1530.946092265625 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 1437.4951969140625 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 1405.1221304296876 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 1343.4887116015625 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 1290.05945640625 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 1261.2783257421875 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 1188.335002421875 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 1244.6512447265625 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 1145.0465993359376 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 1258.005577734375 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 1194.1361692578125 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 1178.327138203125 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 1151.9809562695314 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 1115.4522632421874 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 1212.5351327734375 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 1236.7264633203124 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 1129.7096988671874 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 1124.5999420703124 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 1214.3487353125 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 1097.898828359375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 1076.2484835546875 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 1247.9283218359376 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 1214.075158125 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 1198.93064765625 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 1142.9898784375 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 1205.919777890625 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 1262.817801171875 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 1210.975372265625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 1231.7331519726563 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 1328.930333359375 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 1254.761289140625 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 1125.5186128125 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 1262.8685775195313 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 1237.459378671875 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 1204.0073876953124 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 1137.405095078125 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 1258.9069240625 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 1275.553220390625 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 1346.54895 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 1248.5952147265625 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 1208.6872989648436 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 1295.82844421875 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 1316.0796097070313 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 432.080549765625 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 381.3125614453125 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 359.7443784375 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 330.96993841796876 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 320.38391126953127 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 299.92142598632813 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 292.0546185205078 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 283.10320901367186 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 273.0513530078125 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 267.9574522070312 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 262.1898098388672 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 251.405320703125 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 245.4019890673828 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 240.36019327148438 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 235.46463319824218 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 230.54692152832033 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 232.6379735986328 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 225.51191487304686 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 216.49266020507812 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 213.13386258789063 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 217.53731814941406 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 208.73003357421874 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 206.97497490234375 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 210.77752537109376 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 195.20160494140626 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 200.7779848046875 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 201.7726231640625 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 194.9952972314453 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 196.31784922851563 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 194.3363351171875 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 196.57610537597657 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 187.25579179199218 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 186.91551185546874 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 183.84816095214845 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 183.88843528320314 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 189.78843431640624 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 187.7555524658203 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 185.55689125976562 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 181.3828788183594 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 179.399421640625 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 179.9919827246094 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 176.98932411865235 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 180.73735264648437 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 183.74677614746093 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 179.24562399414063 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 185.12150788574218 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 179.334324765625 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 171.57741375 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 180.4202580859375 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 174.17015067871094 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3521, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 161.696659375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2537912832, + "peak_memory_reserved_bytes": 3323985920, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.146927833557129, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5882.5757439136505, + "train_wall_s": 5881.403960943222 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 30310400000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1407562240000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json b/results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json new file mode 100644 index 0000000..d1663db --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json @@ -0,0 +1,2089 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.00026144811999984086, + "innovation_negative_gradient_cosine": [ + -0.0004968902794644237, + 0.0013761291047558188, + 0.00010523246601223946, + 2.608442446216941e-05, + 0.0005782860680483282, + -0.0003741116961464286, + 0.0006768519524484873, + 0.0011184385512024164, + -0.0004898424376733601, + 9.430304635316133e-05, + -0.0013928385451436043, + 0.0001874021254479885, + -0.000301648979075253, + 0.0015553411794826388, + 0.0013286168687045574, + -0.00010815382120199502, + 0.004846717230975628, + -0.0006915669655427337, + 0.0048540569841861725, + -0.000358888937626034, + -0.0011774199083447456, + 0.007732397876679897, + 0.009962676092982292, + 0.016084017232060432, + 0.011626636609435081, + 0.012131083756685257, + 0.014776289463043213, + 0.007380135823041201, + 0.015810808166861534, + 0.020616794005036354, + 0.025164347141981125 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0004968902794644237, + 0.0013761291047558188, + 0.00010523246601223946, + 2.608442446216941e-05, + 0.0005782860680483282, + -0.0003741116961464286, + 0.0006768519524484873, + 0.0011184385512024164, + -0.0004898424376733601, + 9.430304635316133e-05, + -0.0013928385451436043, + 0.0001874021254479885, + -0.000301648979075253, + 0.0015553411794826388, + 0.0013286168687045574, + -0.00010815382120199502, + 0.004846717230975628, + -0.0006915669655427337, + 0.0048540569841861725, + -0.000358888937626034, + -0.0011774199083447456, + 0.007732397876679897, + 0.009962676092982292, + 0.016084017232060432, + 0.011626636609435081, + 0.012131083756685257, + 0.014776289463043213, + 0.007380135823041201, + 0.015810808166861534, + 0.020616794005036354, + 0.025164347141981125 + ], + "teaching_negative_gradient_cosine": [ + -0.0004968902794644237, + 0.0013761291047558188, + 0.00010523246601223946, + 2.608442446216941e-05, + 0.0005782860680483282, + -0.0003741116961464286, + 0.0006768519524484873, + 0.0011184385512024164, + -0.0004898424376733601, + 9.430304635316133e-05, + -0.0013928385451436043, + 0.0001874021254479885, + -0.000301648979075253, + 0.0015553411794826388, + 0.0013286168687045574, + -0.00010815382120199502, + 0.004846717230975628, + -0.0006915669655427337, + 0.0048540569841861725, + -0.000358888937626034, + -0.0011774199083447456, + 0.007732397876679897, + 0.009962676092982292, + 0.016084017232060432, + 0.011626636609435081, + 0.012131083756685257, + 0.014776289463043213, + 0.007380135823041201, + 0.015810808166861534, + 0.020616794005036354, + 0.025164347141981125 + ], + "wall_s": 0.12825512886047363 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 9.430248809814453 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 9.950192311401366 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 12.234537961730958 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 16.58897698913574 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 22.226687786865234 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 30.449551181640626 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 36.34107797546387 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 46.41980688720703 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 64.11765653808594 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 82.97689662597656 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 102.98532927246094 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 120.77116177001953 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 140.28046770751953 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 157.43301540039062 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 205.60902872070312 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 230.8937860522461 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 253.97255029785157 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 317.77819536132813 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 343.6580914453125 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 407.3898486328125 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 477.45794138671874 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 524.5825004492187 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 557.1729856640625 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 662.8630315625 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 652.32735375 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 747.8931953320313 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 763.5256337109375 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 835.3264582617187 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 1058.0156810351561 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 905.8362717578125 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 1242.1385985546874 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 1134.6342613671875 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 1470.7842817578126 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 1540.7826381640625 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 1596.2331265625 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 1612.490711953125 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 1783.94997171875 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 1837.744021875 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 2026.28462453125 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 2168.968517890625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 2237.83662625 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 2496.417368984375 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 2410.90841 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 2439.45786546875 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 2617.1114828125 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 2940.740397421875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 2858.628394921875 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 3138.9177421875 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 3438.415418515625 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 3396.234941796875 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 3606.21058734375 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 4012.36167828125 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 4042.69706984375 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 4233.11868765625 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 4537.82246765625 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 4367.230910859375 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 4864.74235125 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 5054.6145271875 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 4842.692507109375 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 5127.00258875 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 4986.25878015625 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 5355.6228540625 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 5600.529760234375 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 5942.63540984375 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 6220.42150453125 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 6801.0142440625 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 6335.4780465625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 6611.8889453125 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 7561.9144178125 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 7706.893026875 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 7683.9504371875 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 8865.1978615625 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 8177.27299953125 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 8050.2239734375 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 8500.7814559375 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 8926.3932909375 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 8798.44918875 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 9649.652635625 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 10129.6659884375 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 9658.81137625 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 10053.70207609375 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 10468.5874915625 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 10511.939255625 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 11405.4140125 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 13573.540273125 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 11169.5221915625 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 12833.8986875 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 12101.8815859375 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 12950.2093375 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 12925.1472246875 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 13137.740811875 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 13952.244633125 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 14272.286441875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 14429.253629375 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 14849.0209059375 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 15120.0954565625 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 15741.8414025 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 16062.643003125 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 16957.89691375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 16931.83966375 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 4130.02909078125 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 2515.256910625 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 2213.87051734375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 2094.392287890625 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 1940.17430265625 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 1783.9225576953124 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 1632.972233046875 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 1545.16946125 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 1479.5323171875 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 1357.0434798046874 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 1282.2471298828125 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 1228.005009921875 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 1256.486622109375 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 1209.0455446875 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 1157.4591321484374 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 1197.2893698046876 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 1168.0829266796875 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 1162.90478453125 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 1268.49559921875 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 1171.266567890625 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 1259.3495566992187 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 1226.919266015625 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 1272.689976875 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 1207.8942051953125 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 1252.519639140625 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 1290.55786078125 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 1228.877055078125 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 1228.168096953125 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 1226.1055788867188 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 1368.034173828125 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 1175.8067020703124 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 1334.3213462109375 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 1247.6840731835937 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 1313.6224965625 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 1221.8752206640625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 1328.462998671875 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 1275.940348515625 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 1252.2436762109376 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 1305.7217723828126 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 1340.4838646484375 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 1228.717749453125 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 1456.4811312109375 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 1288.603296171875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 1299.126255078125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 1265.7983320703124 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 1291.65321625 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 1210.3560226171876 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 1396.2364810546876 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 1282.69460765625 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 1252.808085390625 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 409.71976322265624 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 361.3236141210937 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 336.7964623730469 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 316.91355252929685 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 301.974838515625 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 290.2720501855469 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 280.62712240234373 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 272.9833394970703 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 259.4398801367187 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 249.6847511279297 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 246.0232903857422 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 230.45652196777343 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 225.3271287988281 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 218.68095298828126 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 222.911926796875 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 216.82613666992188 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 206.59793936523437 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 200.18790624023438 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 202.72470991210938 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 199.54272793945313 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 188.69470041992187 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 190.51382822753905 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 188.02758157714842 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 186.09762303710937 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 183.76290793945313 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 181.99154977539064 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 185.6296263623047 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 185.9326528515625 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 173.08992820800782 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 173.0970648486328 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 174.8073498828125 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 173.26389734375 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 178.76224380859375 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 177.17293513183594 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 166.40748666503907 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 173.59480986816405 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 179.32762776367187 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 171.93528721679687 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 169.06438075195314 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 165.37803157226563 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 174.83478853515626 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 170.7552182470703 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 165.30579611816407 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 171.17787514648438 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 164.37082291015625 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 184.00646500976563 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 164.81874224365234 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 165.01198385253906 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 168.083658046875 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 166.8696224951172 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3264, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 131.45226796875, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2537912832, + "peak_memory_reserved_bytes": 3323985920, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.146092176437378, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5907.313760757446, + "train_wall_s": 5906.142598867416 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 30310400000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1407562240000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json b/results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json new file mode 100644 index 0000000..782982f --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json @@ -0,0 +1,2089 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0006295824307017029, + "innovation_negative_gradient_cosine": [ + -0.0017798964399844408, + -0.0002305394154973328, + -0.0003138069296255708, + 0.0005825411644764245, + 0.002634990494698286, + -0.0007164775161072612, + 0.002833605743944645, + -0.0009813582291826606, + 0.0018706656992435455, + 0.0023960997350513935, + -0.001758429454639554, + 0.004448906052857637, + 0.005853395909070969, + 0.004932241979986429, + 0.0018295261543244123, + 0.003428603522479534, + 0.005708863027393818, + 0.0003740524407476187, + 0.00368067086674273, + 0.0057233176194131374, + 0.0038504861295223236, + 0.014013165608048439, + 0.013460400514304638, + 0.010330073535442352, + 0.013087764382362366, + 0.00992564670741558, + 0.013553736731410027, + 0.01602642983198166, + 0.017832670360803604, + 0.01908412203192711, + 0.03177158161997795 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0017798964399844408, + -0.0002305394154973328, + -0.0003138069296255708, + 0.0005825411644764245, + 0.002634990494698286, + -0.0007164775161072612, + 0.002833605743944645, + -0.0009813582291826606, + 0.0018706656992435455, + 0.0023960997350513935, + -0.001758429454639554, + 0.004448906052857637, + 0.005853395909070969, + 0.004932241979986429, + 0.0018295261543244123, + 0.003428603522479534, + 0.005708863027393818, + 0.0003740524407476187, + 0.00368067086674273, + 0.0057233176194131374, + 0.0038504861295223236, + 0.014013165608048439, + 0.013460400514304638, + 0.010330073535442352, + 0.013087764382362366, + 0.00992564670741558, + 0.013553736731410027, + 0.01602642983198166, + 0.017832670360803604, + 0.01908412203192711, + 0.03177158161997795 + ], + "teaching_negative_gradient_cosine": [ + -0.0017798964399844408, + -0.0002305394154973328, + -0.0003138069296255708, + 0.0005825411644764245, + 0.002634990494698286, + -0.0007164775161072612, + 0.002833605743944645, + -0.0009813582291826606, + 0.0018706656992435455, + 0.0023960997350513935, + -0.001758429454639554, + 0.004448906052857637, + 0.005853395909070969, + 0.004932241979986429, + 0.0018295261543244123, + 0.003428603522479534, + 0.005708863027393818, + 0.0003740524407476187, + 0.00368067086674273, + 0.0057233176194131374, + 0.0038504861295223236, + 0.014013165608048439, + 0.013460400514304638, + 0.010330073535442352, + 0.013087764382362366, + 0.00992564670741558, + 0.013553736731410027, + 0.01602642983198166, + 0.017832670360803604, + 0.01908412203192711, + 0.03177158161997795 + ], + "wall_s": 0.133530855178833 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 10.945983673706055 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 9.268972373046875 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 11.631135036621094 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 15.33683402130127 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 23.313101820068358 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 25.09919560180664 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 32.680282828369144 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 45.92108037231446 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 59.816421459960935 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 72.5789655078125 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 96.03700706054687 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 119.57904437988282 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 137.12588658203126 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 161.45996876953126 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 206.0152785546875 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 242.59920668945313 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 274.3646112109375 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 322.8843250341797 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 353.91449166992186 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 425.56319271972654 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 457.8159681835937 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 519.7735980859375 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 576.164127421875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 697.5201409960938 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 769.9046768164062 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 832.7179850195313 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 859.3094276660156 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 995.9832930078124 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 1094.9401513671876 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 1203.578955625 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 1192.4968762890626 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 1330.959121953125 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 1535.65795578125 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 1571.660817265625 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 1668.7655644140625 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 1800.386894921875 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 1840.12381953125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 2142.220775390625 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 2094.68745 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 2175.59486328125 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 2477.9617656640626 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 2538.88201515625 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 2733.2605650390624 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 2910.646749453125 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 3267.03188890625 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 3110.582402890625 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 3444.714680859375 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 3504.92890515625 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 3936.67706125 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 4224.252715625 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 3919.4437746875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 4145.83366421875 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 4586.4608271875 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 5002.82196125 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 5003.635096484375 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 5529.62706953125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 5188.77828671875 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 5482.045545625 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 5559.4013565625 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 5908.454430625 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 5980.538468203125 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 6472.2701846875 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 7121.7553678125 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 6582.68953703125 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 7461.9865875 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 7860.72135984375 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 7357.63736375 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 8490.441175 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 7804.14765578125 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 9004.9752734375 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 8701.8584590625 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 9813.075638125 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 10199.4175478125 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 9049.182855 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 10973.45527125 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 9971.30596546875 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 10112.501990625 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 9919.8541425 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 10888.5804590625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 11364.8928075 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 12158.661168125 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 13187.3304353125 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 12198.78574125 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 13902.523351875 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 13500.95309015625 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 12426.27800390625 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 13973.0024403125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 15644.7464046875 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 14839.85543 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 14755.5351403125 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 14689.6583528125 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 14835.501405625 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 17823.8316465625 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 17181.902283125 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 17498.508451875 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 17212.2398271875 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 17416.0030225 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 18377.93776125 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 18028.026354375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 17851.23628375 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 4473.729016640625 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 2481.37507703125 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 2250.18129578125 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 2149.58493734375 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 1967.60817125 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 1852.376851640625 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 1777.812954453125 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 1674.043941796875 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 1628.418797734375 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 1604.0854262109376 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 1504.91542609375 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 1479.6587116015626 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 1531.1797834375 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 1457.0685587109374 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 1466.0198483984375 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 1427.258835234375 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 1466.809107421875 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 1469.6581989453125 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 1487.5747997265626 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 1354.257206796875 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 1377.6707595703126 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 1522.67504984375 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 1490.97368515625 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 1427.966161015625 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 1320.568961640625 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 1398.082406484375 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 1448.611054453125 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 1373.0002104296875 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 1389.1033404296875 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 1495.4376337890626 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 1485.04863515625 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 1501.30015125 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 1416.8424656640625 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 1354.535560703125 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 1257.2732968359376 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 1552.74855265625 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 1409.460373125 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 1559.767197578125 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 1487.652125703125 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 1543.125969609375 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 1467.4839623046876 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 1560.67693390625 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 1279.8525775 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 1493.00962359375 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 1334.0112649609375 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 1468.115055546875 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 1388.5694150195313 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 1477.0068703125 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 1480.3632398046875 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 1345.044163203125 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 497.901114921875 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 410.2715630078125 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 382.8667728125 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 356.7439740234375 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 338.90097631835937 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 330.863975546875 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 315.0818691552734 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 302.8783972753906 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 291.8472606933594 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 287.8466359472656 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 278.3380192578125 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 269.89768504394533 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 270.5264351953125 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 262.18027779296875 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 248.15325690429688 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 248.11462529296875 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 242.40116031738282 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 240.77935544921874 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 235.85267098632812 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 229.2045337060547 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 228.73335375976563 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 222.9525912207031 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 226.910256953125 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 211.91460568359375 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 220.34502372070313 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 217.32782608398438 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 217.8464500390625 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 213.92846335449218 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 209.2568585546875 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 215.55283462402343 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 215.38840415527343 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 205.83364685058595 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 215.3645288623047 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 207.5909344824219 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 203.62799319335937 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 213.9936000732422 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 202.21468946777344 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 201.94267767089843 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 201.10531641601563 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 198.05497042480468 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 203.1494203613281 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 203.26228122070313 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 206.91613942871095 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 191.9774615234375 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 195.60639663085936 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 203.9393654003906 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 196.9610649609375 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 196.74086037597655 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 198.9787810498047 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 194.3087822363281 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3404, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 160.818975, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2537912832, + "peak_memory_reserved_bytes": 3323985920, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1457910537719727, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5887.7735023498535, + "train_wall_s": 5886.60419011116 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 30310400000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1407562240000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json b/results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json new file mode 100644 index 0000000..08910f4 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json @@ -0,0 +1,2089 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0007135697436751798, + "innovation_negative_gradient_cosine": [ + 0.0006930246017873287, + 0.00025351723888888955, + -0.0003096466534771025, + 0.00023615991813130677, + 0.0005241597536951303, + -0.0010539605282247066, + 0.0004080611979588866, + 0.0015929663786664605, + 0.0022438704036176205, + 0.002547545125707984, + 0.0020522314589470625, + -0.0008863675757311285, + -0.0011268185917288065, + 0.004612869583070278, + 0.0021880711428821087, + 0.003208225592970848, + 0.004468095954507589, + 0.0043890392407774925, + 0.0015347551088780165, + 0.00540960393846035, + 0.0036303233355283737, + 0.01276873704046011, + 0.0168010201305151, + 0.01699742302298546, + 0.018613245338201523, + 0.01744714565575123, + 0.01885955035686493, + 0.01784776709973812, + 0.0240741278976202, + 0.01660817861557007, + 0.02493393048644066 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0006930246017873287, + 0.00025351723888888955, + -0.0003096466534771025, + 0.00023615991813130677, + 0.0005241597536951303, + -0.0010539605282247066, + 0.0004080611979588866, + 0.0015929663786664605, + 0.0022438704036176205, + 0.002547545125707984, + 0.0020522314589470625, + -0.0008863675757311285, + -0.0011268185917288065, + 0.004612869583070278, + 0.0021880711428821087, + 0.003208225592970848, + 0.004468095954507589, + 0.0043890392407774925, + 0.0015347551088780165, + 0.00540960393846035, + 0.0036303233355283737, + 0.01276873704046011, + 0.0168010201305151, + 0.01699742302298546, + 0.018613245338201523, + 0.01744714565575123, + 0.01885955035686493, + 0.01784776709973812, + 0.0240741278976202, + 0.01660817861557007, + 0.02493393048644066 + ], + "teaching_negative_gradient_cosine": [ + 0.0006930246017873287, + 0.00025351723888888955, + -0.0003096466534771025, + 0.00023615991813130677, + 0.0005241597536951303, + -0.0010539605282247066, + 0.0004080611979588866, + 0.0015929663786664605, + 0.0022438704036176205, + 0.002547545125707984, + 0.0020522314589470625, + -0.0008863675757311285, + -0.0011268185917288065, + 0.004612869583070278, + 0.0021880711428821087, + 0.003208225592970848, + 0.004468095954507589, + 0.0043890392407774925, + 0.0015347551088780165, + 0.00540960393846035, + 0.0036303233355283737, + 0.01276873704046011, + 0.0168010201305151, + 0.01699742302298546, + 0.018613245338201523, + 0.01744714565575123, + 0.01885955035686493, + 0.01784776709973812, + 0.0240741278976202, + 0.01660817861557007, + 0.02493393048644066 + ], + "wall_s": 0.1259937286376953 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 8.885662679138184 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 8.873351764526367 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 10.523940482177734 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 14.166257419433594 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 17.697643399047852 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 22.953530103149415 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 33.861193767089844 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 43.3194745703125 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 54.50201330810547 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 71.80728986572265 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 84.83739865234375 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 116.18452405517579 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 124.06628524902344 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 138.6655530493164 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 179.30228564941407 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 202.8265821875 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 258.8684362011719 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 246.25785841308593 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 324.236358984375 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 376.75855177734377 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 447.2875923535156 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 464.5191718359375 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 482.16474328125 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 585.5187628710937 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 692.00428421875 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 681.547618046875 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 825.729226953125 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 824.1751372460938 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 958.9309803320313 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 1065.8346053710939 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 1085.8403963671874 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 1181.5517649609376 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 1312.19564390625 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 1427.726493984375 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 1616.18884859375 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 1858.750594609375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 1807.61567953125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 1790.74209265625 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 1975.14533515625 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 2349.380251484375 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 2143.0995401171876 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 2357.918343125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 2556.11945171875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 2527.44374125 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 2824.963735 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 3050.7419096875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 2988.132359765625 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 3266.4915703125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 3583.411778125 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 3418.094278125 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 3615.1833865625 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 4074.670904375 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 4005.102707578125 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 4052.14866265625 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 4113.71191921875 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 4656.29436578125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 4656.771618515625 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 5151.35133859375 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 5102.476401171875 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 5173.2870353125 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 5772.5653415625 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 5678.9386259375 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 6244.67910125 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 6328.75233984375 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 6564.571815625 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 6927.61532125 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 7323.84527 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 6918.079596875 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 8083.695476875 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 7319.2224225 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 8799.09626375 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 8243.33843375 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 9431.78561390625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 8274.32926125 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 9977.9765209375 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 9190.4313346875 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 9297.16496578125 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 9501.1392225 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 10224.1858175 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 9808.345829375 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 10950.676543125 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 10793.331165625 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 10753.8077559375 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 11647.875028125 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 10658.6340084375 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 11020.818275 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 11799.7495428125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 12713.84546875 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 11942.6876278125 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 12927.948485 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 12512.52836125 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 15423.95205234375 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 14134.58731625 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 14506.948000625 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 13866.099086875 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 13612.6851990625 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 15303.8284709375 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 15936.346960625 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 17292.638946875 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 15361.335605625 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 3676.14055984375 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 2522.29870546875 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 2229.5096124609377 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 1996.593728828125 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 1834.0741719140624 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 1697.5649187109375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 1596.7197078125 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 1525.304521328125 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 1435.291812734375 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 1351.258381328125 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 1353.9311638671875 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 1343.518666953125 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 1229.7389730859375 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 1295.62979515625 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 1245.0034944140625 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 1209.270209140625 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 1227.6047748828125 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 1185.1334834765626 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 1164.7048128515626 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 1205.078364296875 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 1218.0545451171874 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 1224.5220325 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 1147.2890941796875 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 1238.18367921875 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 1259.841901875 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 1158.6707849609375 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 1214.9299673632813 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 1216.136162890625 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 1050.0023735742188 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 1107.405067890625 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 1290.2015492382814 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 1164.56899234375 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 1114.820466171875 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 1163.6818910546874 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 1163.4179517578125 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 1300.3529144921874 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 1228.5953010546875 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 1156.040135859375 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 1103.5443741796876 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 1178.8367403515624 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 1208.8990166210938 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 1097.449611171875 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 1271.9026541796875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 1110.8728445703125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 1289.5882077734375 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 1253.7520938671876 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 1224.7363398046875 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 1056.43241515625 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 1340.4110383007812 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 1105.03471171875 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 386.68540423828125 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 334.39591255859375 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 313.3769979003906 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 289.9292345800781 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 275.50476001953126 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 264.6360742675781 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 253.27556083007812 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 241.3612371484375 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 228.4257856933594 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 219.23440580566407 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 214.07328321777345 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 206.0354009375 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 201.9939564453125 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 194.39837044433594 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 191.2468030761719 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 186.10323663574218 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 179.8349450683594 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 178.56441596191405 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 175.0047825366211 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 169.16444389648439 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 174.33617260253905 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 164.6393528076172 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 162.44749002441407 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 167.82198079589844 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 167.80451487304688 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 164.29974818847657 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 164.52811025878907 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 162.1312367163086 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 157.83586200195313 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 155.45913883789063 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 166.92303909179688 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 159.72698168457032 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 148.46945220458986 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 155.47847979492187 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 154.63424322265624 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 150.73987834960937 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 156.83884181152345 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 147.17977385742188 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 151.08661865722655 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 145.43321890625 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 156.03836974121094 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 147.9519510644531 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 145.4378096826172 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 146.60984659179687 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 152.19081804931642 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 144.10472280761718 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 151.817136953125 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 141.34361458496093 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 147.52483091308594 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 138.9903475390625 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.2863, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 162.33229375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2537912832, + "peak_memory_reserved_bytes": 3323985920, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.145874261856079, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5897.761346578598, + "train_wall_s": 5896.593955278397 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 30310400000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1407562240000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json b/results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json new file mode 100644 index 0000000..f473208 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json @@ -0,0 +1,2089 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0006895592523505911, + "innovation_negative_gradient_cosine": [ + 0.0004034606972709298, + -0.0010629461612552404, + 0.0021686870604753494, + 0.0020202791783958673, + -0.0007853569695726037, + 0.0015421448042616248, + 0.0006243109237402678, + 0.0008940756088122725, + 0.00011606185580603778, + 0.0009748755255714059, + -0.00011973280925303698, + 0.0037723020650446415, + -0.0006894980324432254, + 0.001309990300796926, + 0.004070796072483063, + 0.00426744157448411, + 0.004435792565345764, + 0.002416555304080248, + 0.00285190436989069, + 0.0014083355199545622, + 0.005424518138170242, + 0.011612243950366974, + 0.01848183199763298, + 0.013608722947537899, + 0.018342353403568268, + 0.01284860447049141, + 0.01176365278661251, + 0.014326085336506367, + 0.0171220563352108, + 0.016334880143404007, + 0.020972561091184616 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0004034606972709298, + -0.0010629461612552404, + 0.0021686870604753494, + 0.0020202791783958673, + -0.0007853569695726037, + 0.0015421448042616248, + 0.0006243109237402678, + 0.0008940756088122725, + 0.00011606185580603778, + 0.0009748755255714059, + -0.00011973280925303698, + 0.0037723020650446415, + -0.0006894980324432254, + 0.001309990300796926, + 0.004070796072483063, + 0.00426744157448411, + 0.004435792565345764, + 0.002416555304080248, + 0.00285190436989069, + 0.0014083355199545622, + 0.005424518138170242, + 0.011612243950366974, + 0.01848183199763298, + 0.013608722947537899, + 0.018342353403568268, + 0.01284860447049141, + 0.01176365278661251, + 0.014326085336506367, + 0.0171220563352108, + 0.016334880143404007, + 0.020972561091184616 + ], + "teaching_negative_gradient_cosine": [ + 0.0004034606972709298, + -0.0010629461612552404, + 0.0021686870604753494, + 0.0020202791783958673, + -0.0007853569695726037, + 0.0015421448042616248, + 0.0006243109237402678, + 0.0008940756088122725, + 0.00011606185580603778, + 0.0009748755255714059, + -0.00011973280925303698, + 0.0037723020650446415, + -0.0006894980324432254, + 0.001309990300796926, + 0.004070796072483063, + 0.00426744157448411, + 0.004435792565345764, + 0.002416555304080248, + 0.00285190436989069, + 0.0014083355199545622, + 0.005424518138170242, + 0.011612243950366974, + 0.01848183199763298, + 0.013608722947537899, + 0.018342353403568268, + 0.01284860447049141, + 0.01176365278661251, + 0.014326085336506367, + 0.0171220563352108, + 0.016334880143404007, + 0.020972561091184616 + ], + "wall_s": 0.14098811149597168 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 8.805620474395752 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 8.350835408325196 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 10.681710851745606 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 16.77335671142578 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 22.449754643554687 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 29.35107404296875 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 39.90004733276367 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 51.22556348876953 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 67.13049506347656 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 87.19024083984375 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 102.5911556640625 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 142.09606771240234 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 150.49670286132812 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 187.66721022949218 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 232.1009699267578 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 259.5194998803711 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 301.6947142578125 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 347.43199337890627 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 380.5026979296875 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 440.2815456640625 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 480.6986183984375 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 620.7739518359375 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 662.05214890625 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 702.1567390625 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 728.9709217578124 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 914.5146899804688 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 919.4704205078125 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 1006.300039375 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 1141.4923464453125 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 1232.8432301953126 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 1389.3752531640625 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 1505.9793 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 1456.09502546875 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 1740.0200769140624 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 1597.0995358203124 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 1843.1855471484375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 1895.570485859375 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 2192.7776669921873 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 2443.0740737109377 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 2319.47770984375 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 2628.4377153125 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 2509.300444453125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 2924.14329234375 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 2967.7656375 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 3037.92749359375 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 3441.17554046875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 3786.376461796875 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 3879.262553203125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 3987.2079546875 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 4180.7807909375 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 4283.356380625 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 4306.2200528125 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 4606.843755859375 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 5018.36130515625 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 5316.54833703125 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 4643.0150496875 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 5808.17254265625 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 5542.2745303125 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 6338.4386734375 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 6154.35571515625 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 6259.9624228125 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 6753.149336953125 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 6818.915559375 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 7721.2591634375 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 7139.191845625 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 7334.911221484375 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 8110.1671503125 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 9028.808891875 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 8407.42225625 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 8720.909369921876 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 9033.7398475 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 10016.533121875 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 9334.79095625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 10243.53509625 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 10139.2392896875 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 10948.3603975 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 11674.938049375 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 12539.0747546875 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 11396.5133809375 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 12084.41676875 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 10146.2811603125 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 13151.83974875 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 13753.3481084375 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 12972.7673415625 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 15343.30547625 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 12606.09071125 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 14355.98277125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 16031.870379375 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 15331.95300375 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 14208.61420625 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 15304.1598978125 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 17639.0484359375 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 15641.3109246875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 17438.4967225 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 16683.18227875 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 17681.132186875 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 17997.267080625 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 16908.620541875 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 20111.61025625 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 19823.012898125 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 4496.975081328125 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 3077.813718125 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 2625.959957109375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 2376.72200609375 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 2211.7159503125 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 2075.416881953125 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 1950.8858953515626 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 1840.804065390625 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 1785.536086171875 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 1652.551065390625 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 1694.081301171875 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 1608.796488359375 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 1496.587931640625 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 1468.904155390625 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 1437.1500469921875 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 1501.219335546875 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 1445.367843828125 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 1423.089146171875 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 1452.4740369921874 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 1512.943267734375 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 1490.286588671875 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 1506.9202340234374 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 1346.818848203125 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 1418.061697890625 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 1494.785019765625 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 1382.9908186328125 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 1471.1115754296875 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 1595.224856875 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 1377.9502681445313 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 1369.924039609375 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 1560.395565703125 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 1474.6887964453124 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 1404.60084453125 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 1345.7678756640626 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 1404.53574296875 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 1506.561594296875 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 1452.3612709765625 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 1402.8059835546876 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 1493.02154453125 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 1525.948107578125 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 1602.414221484375 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 1353.724499375 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 1567.0157021875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 1637.27489265625 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 1520.4823464648437 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 1415.2248896679687 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 1464.837591015625 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 1325.2173388671874 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 1423.434475234375 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 1669.24590078125 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 542.0815853125 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 454.12697358398435 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 423.79257350585937 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 395.2855218457031 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 374.61574724609375 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 353.32607408203125 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 342.70486810546873 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 327.0183475878906 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 318.72834114257813 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 309.9834883691406 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 298.908067265625 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 293.0573503125 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 285.34418674804687 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 280.08773247070314 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 274.38740713867185 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 270.844878203125 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 263.73946388671874 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 257.57104933105467 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 256.59306181640625 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 250.76854859375 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 247.958570625 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 242.4240597558594 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 245.42694532226562 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 245.9434549121094 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 237.9543085058594 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 237.430843984375 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 239.2065728857422 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 234.32829400390625 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 234.61293728515625 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 238.9337918066406 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 230.09997089355468 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 226.23547713867188 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 226.47352134277344 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 234.67925583007812 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 229.8854143017578 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 220.98700480957032 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 229.81571215820313 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 220.7603815136719 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 222.98204373046875 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 218.0057065234375 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 216.22857752929687 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 235.38879555664062 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 213.90988830078126 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 219.10026682617186 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 215.59104669921874 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 211.9366745703125 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 217.60682508789063 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 209.33322073730469 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 212.92625396484374 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 215.82442040039064 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.329, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 191.6427796875, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2537912832, + "peak_memory_reserved_bytes": 3323985920, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1455142498016357, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 5920.709708213806, + "train_wall_s": 5919.539194345474 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 30310400000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 1407562240000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json b/results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json new file mode 100644 index 0000000..e9f577f --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json @@ -0,0 +1,2281 @@ +{ + "architecture": { + "adaptive_apical_parameters": 6389760, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 5324800 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 10, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.00022253473434830084, + "innovation_negative_gradient_cosine": [ + -0.0007358535658568144, + 1.0973031749017537e-05, + 0.00028803618624806404, + 0.0018082298338413239, + 0.0017930353060364723, + 0.0026586810126900673, + -0.0005942065035924315, + -0.00015428336337208748, + -0.0017858229111880064, + -0.0009746977593749762, + 0.0017651284579187632, + -0.0018736472120508552, + 0.0012291024904698133, + -0.0015807488234713674, + 0.0010712535586208105, + -0.0008963914588093758, + 0.0010419270256534219, + 0.0009349099127575755, + 0.0007942283409647644, + 0.0022612982429564, + 0.0029903706163167953, + -0.0006342272972688079, + -0.0012854530941694975, + 0.0017757457681000233, + 0.0026695728302001953, + 0.0005210742820054293, + 0.0027406071312725544, + 0.003057739231735468, + 0.004302687011659145, + -0.0002525838790461421, + 0.00313854543492198, + 0.0034365016035735607, + 0.0020225904881954193, + 0.0012177969329059124, + 0.0021197060123085976, + 0.0009309356100857258, + 0.0001798642915673554, + 0.006009994074702263, + 0.00971902348101139, + 0.011476058512926102, + 0.011707705445587635, + 0.0097414031624794, + 0.014721017330884933, + 0.01197153888642788, + 0.010873015969991684, + 0.013532187789678574, + 0.017107289284467697, + 0.013830152340233326, + 0.019055891782045364, + 0.009968715719878674, + 0.011132730171084404, + 0.015165727585554123, + 0.013714831322431564, + 0.014137480407953262, + 0.016913186758756638 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0007358535658568144, + 1.0973031749017537e-05, + 0.00028803618624806404, + 0.0018082298338413239, + 0.0017930353060364723, + 0.0026586810126900673, + -0.0005942065035924315, + -0.00015428336337208748, + -0.0017858229111880064, + -0.0009746977593749762, + 0.0017651284579187632, + -0.0018736472120508552, + 0.0012291024904698133, + -0.0015807488234713674, + 0.0010712535586208105, + -0.0008963914588093758, + 0.0010419270256534219, + 0.0009349099127575755, + 0.0007942283409647644, + 0.0022612982429564, + 0.0029903706163167953, + -0.0006342272972688079, + -0.0012854530941694975, + 0.0017757457681000233, + 0.0026695728302001953, + 0.0005210742820054293, + 0.0027406071312725544, + 0.003057739231735468, + 0.004302687011659145, + -0.0002525838790461421, + 0.00313854543492198, + 0.0034365016035735607, + 0.0020225904881954193, + 0.0012177969329059124, + 0.0021197060123085976, + 0.0009309356100857258, + 0.0001798642915673554, + 0.006009994074702263, + 0.00971902348101139, + 0.011476058512926102, + 0.011707705445587635, + 0.0097414031624794, + 0.014721017330884933, + 0.01197153888642788, + 0.010873015969991684, + 0.013532187789678574, + 0.017107289284467697, + 0.013830152340233326, + 0.019055891782045364, + 0.009968715719878674, + 0.011132730171084404, + 0.015165727585554123, + 0.013714831322431564, + 0.014137480407953262, + 0.016913186758756638 + ], + "teaching_negative_gradient_cosine": [ + -0.0007358535658568144, + 1.0973031749017537e-05, + 0.00028803618624806404, + 0.0018082298338413239, + 0.0017930353060364723, + 0.0026586810126900673, + -0.0005942065035924315, + -0.00015428336337208748, + -0.0017858229111880064, + -0.0009746977593749762, + 0.0017651284579187632, + -0.0018736472120508552, + 0.0012291024904698133, + -0.0015807488234713674, + 0.0010712535586208105, + -0.0008963914588093758, + 0.0010419270256534219, + 0.0009349099127575755, + 0.0007942283409647644, + 0.0022612982429564, + 0.0029903706163167953, + -0.0006342272972688079, + -0.0012854530941694975, + 0.0017757457681000233, + 0.0026695728302001953, + 0.0005210742820054293, + 0.0027406071312725544, + 0.003057739231735468, + 0.004302687011659145, + -0.0002525838790461421, + 0.00313854543492198, + 0.0034365016035735607, + 0.0020225904881954193, + 0.0012177969329059124, + 0.0021197060123085976, + 0.0009309356100857258, + 0.0001798642915673554, + 0.006009994074702263, + 0.00971902348101139, + 0.011476058512926102, + 0.011707705445587635, + 0.0097414031624794, + 0.014721017330884933, + 0.01197153888642788, + 0.010873015969991684, + 0.013532187789678574, + 0.017107289284467697, + 0.013830152340233326, + 0.019055891782045364, + 0.009968715719878674, + 0.011132730171084404, + 0.015165727585554123, + 0.013714831322431564, + 0.014137480407953262, + 0.016913186758756638 + ], + "wall_s": 0.20293188095092773 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 15.54466657989502 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 14.547235075073242 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 17.414953399658202 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 25.90386300109863 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 39.68947647094726 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 54.04766418945312 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 72.88010504882813 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 108.43115664550781 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 154.87382209960938 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 168.45801689208983 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 252.97475659179688 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 292.0759301953125 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 332.7507882080078 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 442.96318801757815 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 564.6440700390625 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 572.4222674023438 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 840.05808921875 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 865.4898714257813 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 1098.5925456054688 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 1292.8258465625 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 1583.3709852734376 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 1701.3120577734376 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 2075.83830796875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 2293.434503203125 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 2558.8233353125 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 3020.405894765625 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 3351.99866171875 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 3603.296686875 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 4202.71841546875 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 4266.887469375 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 4414.28546203125 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 4593.823394375 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 5191.17040109375 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 5839.81197625 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 6529.7030828125 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 6368.560588125 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 7541.834451015625 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 7486.9357040625 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 8868.19930328125 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 9579.1515990625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 9371.672059375 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 10813.829093125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 10746.49713625 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 12715.11721 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 12201.307511875 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 12907.2360246875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 13705.78240875 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 13940.5551475 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 15038.75103625 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 15487.918846875 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 15641.1139175 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 17282.7993675 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 18200.64971375 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 18207.58564625 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 21211.5489225 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 21569.2429675 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 20786.201261875 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 22278.53631875 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 25304.68420375 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 22680.234508125 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 25914.68491625 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 23488.31530125 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 26916.453524375 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 27199.517476875 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 29014.880584375 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 31759.2104296875 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 32718.88177625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 30877.23190875 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 33513.7631775 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 34085.73262625 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 34196.8135575 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 33159.024895 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 37637.4728825 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 36281.769501875 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 39038.17320375 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 37316.9266325 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 41881.991545 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 39957.9294975 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 42248.13161625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 45955.48953875 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 42655.75375875 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 49212.42627 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 53806.118375625 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 53437.919545 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 56606.1347175 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 49374.35269125 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 51668.7037975 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 48634.9196875 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 58801.28231625 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 57543.91696875 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 57933.4511 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 55352.67518375 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 58914.55379 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 65179.508695 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 58191.2278625 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 66903.70380875 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 69500.40155 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 65830.5844625 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 70910.9999825 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 78901.0362975 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 13774.071160625 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 10134.08951875 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 9220.26762609375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 8444.07215015625 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 8040.99920953125 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 7679.8396934375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 7400.2494 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 7037.24590375 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 6829.8353040625 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 6355.3798890625 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 5975.3924253125 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 5699.65335578125 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 5832.3879453125 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 5617.43853234375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 5341.032689375 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 5310.4768853125 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 5184.67361703125 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 5230.7552275 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 5096.06707984375 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 5188.968658125 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 5028.6196609375 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 5261.3139678125 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 4988.89028703125 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 5692.444181875 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 5410.727960625 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 4953.4976396875 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 5121.6054525 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 5625.810119375 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 5545.74374234375 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 5527.4545828125 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 5033.5081665625 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 5074.50682484375 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 5297.9041115625 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 5031.92348703125 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 5217.82683203125 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 5220.15557703125 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 5376.6422384375 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 5205.66982625 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 5085.99358359375 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 5418.01852515625 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 5023.23353546875 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 5162.695340078125 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 5160.7980140625 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 5663.014644375 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 5296.452984296875 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 5229.27388703125 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 5626.622417578125 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 5444.78413109375 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 5456.180879921875 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 5576.560659453125 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 1782.5386803515626 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 1514.4121122265626 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 1410.9323049609375 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 1311.8951089453126 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 1214.2603724609376 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 1161.9654675 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 1120.774872265625 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 1080.1324510546874 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 1043.7089065625 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 1012.7494857421875 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 980.0301846875 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 946.735686796875 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 946.5751611132813 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 914.8629967578125 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 914.1067816796875 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 912.10888921875 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 884.2077573828125 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 902.1880673632812 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 878.38894671875 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 833.30157421875 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 866.5463003125 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 855.2307503515625 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 835.0665584375 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 841.1369849804687 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 821.0857187304688 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 815.5904630078124 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 829.5957598242187 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 815.3975285742188 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 828.9468991796875 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 827.04718890625 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 823.0365430078125 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 813.0535859179688 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 817.4472450195312 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 801.30095453125 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 796.7622555859375 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 789.8825111328125 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 793.7847402929688 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 779.5417561328125 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 793.7602258203125 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 806.3877822070313 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 797.7902808984375 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 755.6947116015625 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 779.7081112304687 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 803.3361581640625 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 772.281200703125 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 782.7138865039062 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 780.0067841015625 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 778.2932036523438 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 756.139004296875 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 788.0550848046875 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3348, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 831.0552375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3339442688, + "peak_memory_reserved_bytes": 4676648960, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.034506320953369, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10377.251267910004, + "train_wall_s": 10375.189896583557 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 53248000000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2562961920000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json b/results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json new file mode 100644 index 0000000..3dff1aa --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json @@ -0,0 +1,2281 @@ +{ + "architecture": { + "adaptive_apical_parameters": 6389760, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 5324800 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 11, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0005879553298452973, + "innovation_negative_gradient_cosine": [ + 0.0011305890511721373, + -8.942060230765492e-05, + -0.0006630357238464057, + -0.000659326498862356, + -0.0011131414212286472, + 0.0027757694479078054, + 0.002825309056788683, + 0.0013577521312981844, + 0.0004187228623777628, + 0.000324989203363657, + 0.0002259696484543383, + 0.002006225287914276, + -0.0006687602144666016, + -2.0834908355027437e-05, + 9.226246038451791e-05, + 0.000342094455845654, + 0.0011339447228237987, + 0.0011640869779512286, + 0.0015215585008263588, + 0.003560976590961218, + 0.0063767931424081326, + 5.662994226440787e-05, + 0.0038221392314881086, + 0.0007309076609089971, + -0.0009009235654957592, + 0.0029084018897265196, + 0.0029845749959349632, + 0.002053938340395689, + 0.001768512767739594, + 0.003022806253284216, + 0.0038056422490626574, + 0.002493410836905241, + 0.0030819051899015903, + 0.005456170998513699, + 0.0029835072346031666, + 0.004010680131614208, + 0.0027565720956772566, + 0.01105463970452547, + 0.01161639578640461, + 0.009767036885023117, + 0.006078685168176889, + 0.010227176360785961, + 0.007261885330080986, + 0.008559858426451683, + 0.015124209225177765, + 0.0069428132846951485, + 0.010973193682730198, + 0.004366468172520399, + 0.009382097981870174, + 0.008147720247507095, + 0.010429740883409977, + 0.005681689362972975, + 0.015391110442578793, + 0.014982243999838829, + 0.02180664986371994 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0011305890511721373, + -8.942060230765492e-05, + -0.0006630357238464057, + -0.000659326498862356, + -0.0011131414212286472, + 0.0027757694479078054, + 0.002825309056788683, + 0.0013577521312981844, + 0.0004187228623777628, + 0.000324989203363657, + 0.0002259696484543383, + 0.002006225287914276, + -0.0006687602144666016, + -2.0834908355027437e-05, + 9.226246038451791e-05, + 0.000342094455845654, + 0.0011339447228237987, + 0.0011640869779512286, + 0.0015215585008263588, + 0.003560976590961218, + 0.0063767931424081326, + 5.662994226440787e-05, + 0.0038221392314881086, + 0.0007309076609089971, + -0.0009009235654957592, + 0.0029084018897265196, + 0.0029845749959349632, + 0.002053938340395689, + 0.001768512767739594, + 0.003022806253284216, + 0.0038056422490626574, + 0.002493410836905241, + 0.0030819051899015903, + 0.005456170998513699, + 0.0029835072346031666, + 0.004010680131614208, + 0.0027565720956772566, + 0.01105463970452547, + 0.01161639578640461, + 0.009767036885023117, + 0.006078685168176889, + 0.010227176360785961, + 0.007261885330080986, + 0.008559858426451683, + 0.015124209225177765, + 0.0069428132846951485, + 0.010973193682730198, + 0.004366468172520399, + 0.009382097981870174, + 0.008147720247507095, + 0.010429740883409977, + 0.005681689362972975, + 0.015391110442578793, + 0.014982243999838829, + 0.02180664986371994 + ], + "teaching_negative_gradient_cosine": [ + 0.0011305890511721373, + -8.942060230765492e-05, + -0.0006630357238464057, + -0.000659326498862356, + -0.0011131414212286472, + 0.0027757694479078054, + 0.002825309056788683, + 0.0013577521312981844, + 0.0004187228623777628, + 0.000324989203363657, + 0.0002259696484543383, + 0.002006225287914276, + -0.0006687602144666016, + -2.0834908355027437e-05, + 9.226246038451791e-05, + 0.000342094455845654, + 0.0011339447228237987, + 0.0011640869779512286, + 0.0015215585008263588, + 0.003560976590961218, + 0.0063767931424081326, + 5.662994226440787e-05, + 0.0038221392314881086, + 0.0007309076609089971, + -0.0009009235654957592, + 0.0029084018897265196, + 0.0029845749959349632, + 0.002053938340395689, + 0.001768512767739594, + 0.003022806253284216, + 0.0038056422490626574, + 0.002493410836905241, + 0.0030819051899015903, + 0.005456170998513699, + 0.0029835072346031666, + 0.004010680131614208, + 0.0027565720956772566, + 0.01105463970452547, + 0.01161639578640461, + 0.009767036885023117, + 0.006078685168176889, + 0.010227176360785961, + 0.007261885330080986, + 0.008559858426451683, + 0.015124209225177765, + 0.0069428132846951485, + 0.010973193682730198, + 0.004366468172520399, + 0.009382097981870174, + 0.008147720247507095, + 0.010429740883409977, + 0.005681689362972975, + 0.015391110442578793, + 0.014982243999838829, + 0.02180664986371994 + ], + "wall_s": 0.21522068977355957 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 21.822016004943848 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 17.824559099121092 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 21.810150268554686 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 30.373354115600584 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 37.876012268066404 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 53.92637309448242 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 67.34847674072266 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 100.33222059082031 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 133.74985802246093 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 173.3882081298828 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 226.11262048339844 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 290.80861913085937 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 415.3767459765625 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 466.00302681640625 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 575.5295325 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 694.2015186328125 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 922.064875 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 1099.0480919140625 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 1343.856710390625 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 1639.4419911328125 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 1726.9874137890624 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 1736.739409453125 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 2409.0237875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 2551.986870078125 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 2991.640676875 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 3164.708963671875 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 3593.382035234375 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 3886.67451828125 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 4288.502298125 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 4229.5054853125 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 4708.4159971875 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 5456.21266015625 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 6402.48859484375 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 6137.558204375 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 6842.644431875 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 7860.1643009375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 7504.49326 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 8271.88913 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 8202.785735625 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 9878.64996890625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 10267.02852890625 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 11784.039109375 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 11671.2003621875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 11453.28779875 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 13273.710763125 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 12479.77634875 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 13303.353846875 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 14392.4950259375 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 16243.511494375 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 16262.6872675 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 18072.855780625 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 17862.5171825 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 19925.392656875 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 20454.8466 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 21781.7770725 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 21613.798714375 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 22883.51164625 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 22273.373185 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 25506.16983875 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 25642.85067625 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 24691.8515728125 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 25866.07558625 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 29144.162578125 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 27740.450170625 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 29032.3199528125 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 28706.324155 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 29711.05745 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 31706.905618125 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 36933.063976875 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 37287.953715 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 37194.62840875 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 36731.81285375 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 41938.34304 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 42232.847375625 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 41347.69572 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 40861.88819875 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 44865.75046875 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 43689.4772625 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 46145.1057175 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 50945.485785 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 48024.606185 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 46213.86058 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 48255.666905 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 55334.47507875 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 51214.2301425 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 57531.0495775 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 59073.374505 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 59636.9727625 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 58170.847806875 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 64804.94207 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 65841.375075 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 67606.9351425 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 66196.475085 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 64087.29575625 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 70802.3089975 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 72189.4987 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 77082.48973 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 77423.9751925 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 75377.29657375 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 69896.49668125 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 14645.348516875 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 11066.969068125 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 9616.640616875 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 9555.9286115625 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 8889.608115625 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 8763.56310015625 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 8327.51983515625 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 8319.9284575 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 7766.7889059375 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 7218.4834346875 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 6762.2607203125 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 6680.30279734375 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 6376.58596 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 6162.12700375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 6278.253233046875 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 6274.7127134375 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 5905.41486 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 5748.2783265625 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 6044.13596671875 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 5841.58058859375 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 6033.3200440625 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 5670.89361828125 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 6045.1858071875 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 5674.30254171875 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 6182.4114390625 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 5883.83195515625 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 6481.4023884375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 5811.732153203125 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 5859.4066059375 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 5458.5892265625 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 5791.215790625 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 5406.5100315625 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 5945.68840109375 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 5955.9906925 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 6088.13583 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 5708.5804115625 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 5858.96247203125 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 5559.608916796875 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 6280.13497953125 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 6046.8483821875 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 6157.783931953125 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 5573.92391609375 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 6364.5391053125 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 5800.603029921875 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 5515.605156015625 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 6255.1965696875 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 5524.64677375 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 5814.7748246875 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 5768.510907890625 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 6645.0301459375 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 2053.8047731640627 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 1795.6883098046876 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 1641.6798268359375 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 1529.563673671875 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 1432.554256171875 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 1364.2565076953124 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 1310.4285279882813 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 1249.9394315820311 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 1218.193686953125 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 1164.8607373828124 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 1151.4370148046876 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 1109.2520802734375 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 1062.9285024609376 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 1052.6393740625 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 1033.6278116796875 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 1011.65394265625 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 992.3858441992187 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 960.9110959765625 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 948.3309200390624 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 927.1700156445313 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 909.3944135546875 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 897.6867053515625 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 910.5499123828125 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 894.3773679882812 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 889.3560428515625 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 884.3611259765624 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 867.3764177734375 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 836.3760258203125 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 853.8147291015625 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 850.666523984375 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 878.9937553515625 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 858.25947765625 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 835.6640744921875 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 867.6950784570313 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 887.0921052734375 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 853.1495698828124 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 848.339279765625 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 845.5097470703125 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 840.1051002929687 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 846.5633176953125 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 864.5601016015625 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 823.3140968164063 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 817.0898944335937 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 857.8201413671875 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 818.969621796875 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 808.2602341015624 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 840.3612466796875 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 820.9785159765624 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 807.97784828125 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 882.8125062890625 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.2604, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 835.0514375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3339442688, + "peak_memory_reserved_bytes": 4676648960, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.036393165588379, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10413.29232931137, + "train_wall_s": 10411.230472564697 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 53248000000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2562961920000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json b/results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json new file mode 100644 index 0000000..694bcde --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json @@ -0,0 +1,2281 @@ +{ + "architecture": { + "adaptive_apical_parameters": 6389760, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 5324800 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 12, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.00033107280048878037, + "innovation_negative_gradient_cosine": [ + -0.001333452295511961, + 0.0001211967901326716, + 0.0003148901159875095, + -0.0016955535393208265, + 0.0011136706452816725, + -0.0010295998072251678, + -8.456464274786413e-05, + 1.7847312847152352e-05, + -0.00017168925842270255, + -0.0007532823365181684, + -0.00022904499201104045, + -0.0003966555232182145, + 0.0004654741787817329, + 0.0008271206170320511, + -0.0006747680599801242, + 6.877288979012519e-05, + -0.0014168135821819305, + -0.0011028589215129614, + -0.0001397784799337387, + 0.0001774124102666974, + 0.002049159724265337, + 0.002690684050321579, + 0.0026326305232942104, + 0.0013554745819419622, + -0.000294134602881968, + -0.00032216659747064114, + 0.0016200796235352755, + -0.0013347950298339128, + 0.0017341312486678362, + 0.002476324327290058, + 0.0011687445221468806, + 0.0002642289618961513, + 0.004258967936038971, + 0.0023720012977719307, + -0.00039734109304845333, + -0.00044895545579493046, + 0.002411904511973262, + 0.006323819514364004, + 0.008260108530521393, + 0.004710537847131491, + 0.008059056475758553, + 0.004862384404987097, + 0.00503207603469491, + 0.010193176567554474, + 0.004132219590246677, + 0.008396388962864876, + 0.008834676817059517, + 0.0075888317078351974, + 0.006854957435280085, + 0.008016476407647133, + 0.006668945774435997, + 0.00800873339176178, + 0.010383546352386475, + 0.009535139426589012, + 0.01782611943781376 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.001333452295511961, + 0.0001211967901326716, + 0.0003148901159875095, + -0.0016955535393208265, + 0.0011136706452816725, + -0.0010295998072251678, + -8.456464274786413e-05, + 1.7847312847152352e-05, + -0.00017168925842270255, + -0.0007532823365181684, + -0.00022904499201104045, + -0.0003966555232182145, + 0.0004654741787817329, + 0.0008271206170320511, + -0.0006747680599801242, + 6.877288979012519e-05, + -0.0014168135821819305, + -0.0011028589215129614, + -0.0001397784799337387, + 0.0001774124102666974, + 0.002049159724265337, + 0.002690684050321579, + 0.0026326305232942104, + 0.0013554745819419622, + -0.000294134602881968, + -0.00032216659747064114, + 0.0016200796235352755, + -0.0013347950298339128, + 0.0017341312486678362, + 0.002476324327290058, + 0.0011687445221468806, + 0.0002642289618961513, + 0.004258967936038971, + 0.0023720012977719307, + -0.00039734109304845333, + -0.00044895545579493046, + 0.002411904511973262, + 0.006323819514364004, + 0.008260108530521393, + 0.004710537847131491, + 0.008059056475758553, + 0.004862384404987097, + 0.00503207603469491, + 0.010193176567554474, + 0.004132219590246677, + 0.008396388962864876, + 0.008834676817059517, + 0.0075888317078351974, + 0.006854957435280085, + 0.008016476407647133, + 0.006668945774435997, + 0.00800873339176178, + 0.010383546352386475, + 0.009535139426589012, + 0.01782611943781376 + ], + "teaching_negative_gradient_cosine": [ + -0.001333452295511961, + 0.0001211967901326716, + 0.0003148901159875095, + -0.0016955535393208265, + 0.0011136706452816725, + -0.0010295998072251678, + -8.456464274786413e-05, + 1.7847312847152352e-05, + -0.00017168925842270255, + -0.0007532823365181684, + -0.00022904499201104045, + -0.0003966555232182145, + 0.0004654741787817329, + 0.0008271206170320511, + -0.0006747680599801242, + 6.877288979012519e-05, + -0.0014168135821819305, + -0.0011028589215129614, + -0.0001397784799337387, + 0.0001774124102666974, + 0.002049159724265337, + 0.002690684050321579, + 0.0026326305232942104, + 0.0013554745819419622, + -0.000294134602881968, + -0.00032216659747064114, + 0.0016200796235352755, + -0.0013347950298339128, + 0.0017341312486678362, + 0.002476324327290058, + 0.0011687445221468806, + 0.0002642289618961513, + 0.004258967936038971, + 0.0023720012977719307, + -0.00039734109304845333, + -0.00044895545579493046, + 0.002411904511973262, + 0.006323819514364004, + 0.008260108530521393, + 0.004710537847131491, + 0.008059056475758553, + 0.004862384404987097, + 0.00503207603469491, + 0.010193176567554474, + 0.004132219590246677, + 0.008396388962864876, + 0.008834676817059517, + 0.0075888317078351974, + 0.006854957435280085, + 0.008016476407647133, + 0.006668945774435997, + 0.00800873339176178, + 0.010383546352386475, + 0.009535139426589012, + 0.01782611943781376 + ], + "wall_s": 0.20418858528137207 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 20.94000764160156 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 19.62649717285156 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 25.090193415527345 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 29.649140238647462 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 34.545018566894534 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 48.3627574975586 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 77.06149560791016 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 101.49769426757813 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 150.34198771972657 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 177.8562303808594 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 221.7907431298828 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 294.5412326269531 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 344.3779750390625 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 411.1375525390625 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 560.7590898242188 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 623.2798921875 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 752.5977274804687 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 1002.7074587304687 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 1313.94790640625 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 1414.9425067773439 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 1503.092805625 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 1937.405467265625 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 2087.423003203125 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 2564.388800546875 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 2609.4218796875 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 3110.83165625 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 3776.0829210546876 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 3880.975733046875 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 4387.7525015625 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 4984.768659375 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 5330.03115265625 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 5846.50104453125 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 5710.3653346875 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 6278.23847453125 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 6830.864655 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 6546.9267015625 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 7391.49460296875 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 8033.417875 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 8922.7807009375 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 8632.656355625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 9130.92773734375 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 10076.9714134375 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 11774.00065875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 11593.69164125 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 11887.653268125 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 12574.010313125 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 14421.15337375 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 13520.107968125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 15746.6091353125 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 15733.638566875 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 18329.241820625 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 18431.455229375 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 17570.81183625 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 16973.86219875 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 19579.385968125 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 23204.30533 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 20706.27900125 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 21582.72516875 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 26746.427590625 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 26144.13405625 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 24698.052254375 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 23851.24748625 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 27929.38762125 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 30175.7901 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 28197.3008525 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 29571.4402075 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 31887.56947375 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 31777.02970375 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 33884.86967375 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 39072.53294125 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 37428.5231675 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 36754.15567 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 40484.509706875 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 41803.5729125 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 35832.9667425 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 40203.92737625 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 40912.255264375 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 43315.66097625 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 43469.07302875 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 43828.5124 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 45599.000315 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 40848.565963125 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 50698.45087125 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 45765.55349375 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 47759.3486175 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 47235.41903 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 55139.881675 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 54489.0964925 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 50621.15795 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 52931.92026625 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 61464.17159625 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 58305.407645 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 57352.33292125 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 58644.4164275 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 62408.0420275 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 61532.5740675 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 59007.93284375 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 68775.532 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 64542.84597625 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 70629.09657 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 14654.2372521875 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 10468.6515740625 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 9067.7611084375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 8423.380498125 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 8158.55437375 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 7619.72909359375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 7046.0333971875 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 6775.8592484375 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 6478.0563859375 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 6159.16063015625 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 5799.35488171875 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 5433.1685890625 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 5350.97988546875 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 4986.5760428125 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 5325.34049234375 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 5104.84509546875 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 5064.1554721875 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 5048.13006140625 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 4887.15388859375 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 4995.7025578125 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 5238.9835584375 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 4834.56711734375 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 5015.1136084375 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 5329.49159671875 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 5179.48424609375 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 4766.6652128125 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 5021.14735359375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 5365.06995828125 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 4384.51806296875 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 5167.5980784375 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 5362.407710546875 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 4922.89476328125 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 5145.946841796875 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 4707.67387203125 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 5891.3174746875 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 4795.557129375 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 4844.1970709375 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 5076.74701140625 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 5042.289148828125 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 5365.8960978125 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 5071.8902071875 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 5137.35602296875 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 5353.469021875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 5375.27371984375 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 5149.46602109375 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 5708.603248125 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 5485.7706896875 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 5218.4024771875 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 5457.35446234375 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 5654.4977634375 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 1917.4399200390626 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 1666.58332859375 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 1537.5342199609374 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 1442.5484925390624 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 1339.382270703125 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 1298.7273259375 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 1237.4698423242187 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 1184.8028786328125 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 1164.850845234375 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 1108.2915394140625 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 1075.1936887890624 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 1055.1205018554688 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 1035.362221640625 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 988.9151735546875 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 993.6608714257812 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 960.6094983984375 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 925.7214187695313 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 945.243518203125 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 915.3768917578125 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 900.5765274609375 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 891.4864741992187 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 869.0761085742188 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 884.9851100195312 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 876.4683806445313 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 848.3266902734375 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 845.732801015625 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 839.8744810351562 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 845.5655024609375 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 824.5605144726562 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 834.7904227929688 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 850.2991547460938 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 820.90386671875 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 827.5199787695312 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 832.7114706640625 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 800.0516747265625 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 794.9772009375 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 802.980529453125 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 817.6020956640625 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 798.885476328125 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 825.81932140625 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 789.4225649414062 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 817.1468815625 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 803.5091401757812 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 799.3935744921876 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 781.3792027734376 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 796.2545483984375 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 851.6191300195312 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 756.63056 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 822.309709140625 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 765.2383037109375 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3382, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 606.93005625, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3339442688, + "peak_memory_reserved_bytes": 4676648960, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0357413291931152, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10408.400108575821, + "train_wall_s": 10406.33801150322 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 53248000000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2562961920000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json b/results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json new file mode 100644 index 0000000..24ebf20 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json @@ -0,0 +1,2281 @@ +{ + "architecture": { + "adaptive_apical_parameters": 6389760, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 5324800 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 13, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.00043991863236038224, + "innovation_negative_gradient_cosine": [ + 0.0005255909636616707, + -0.0010781302116811275, + 0.0025069178082048893, + 0.0016919649206101894, + 0.00010788301005959511, + 2.1327403374016285e-05, + -0.000367272412404418, + 0.0003776442026719451, + 0.0016317082336172462, + 0.0014963049907237291, + -0.0005021097604185343, + -0.0002937753451988101, + -0.0004791772225871682, + 6.390659837052226e-05, + -0.001030154642648995, + 0.0014018112560734153, + 0.0018113252008333802, + 3.277038922533393e-05, + -4.067507688887417e-05, + -0.0006657886551693082, + 0.0020185261964797974, + -0.0015718279173597693, + -0.0024743354879319668, + 0.0002824710390996188, + 0.003487984649837017, + 0.0005018783849664032, + -0.0022507067769765854, + -0.0005763396038673818, + 0.0007501582731492817, + 0.0011279437458142638, + 0.0008351888973265886, + 0.0006766966544091702, + 0.0011843631509691477, + -3.760406980291009e-05, + 0.003942208830267191, + 0.00019845661881845444, + 0.002232921775430441, + 0.004017248749732971, + 0.007412542589008808, + 0.0004585348069667816, + 0.007836693897843361, + 0.004988133907318115, + 0.002706740517169237, + 0.003983347676694393, + 0.006801050156354904, + 0.0037632600869983435, + 0.010236972942948341, + 0.00855157058686018, + 0.009376580826938152, + 0.006161993369460106, + 0.008103354834020138, + 0.011376159265637398, + 0.014972970820963383, + 0.012046518735587597, + 0.010956445708870888 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0005255909636616707, + -0.0010781302116811275, + 0.0025069178082048893, + 0.0016919649206101894, + 0.00010788301005959511, + 2.1327403374016285e-05, + -0.000367272412404418, + 0.0003776442026719451, + 0.0016317082336172462, + 0.0014963049907237291, + -0.0005021097604185343, + -0.0002937753451988101, + -0.0004791772225871682, + 6.390659837052226e-05, + -0.001030154642648995, + 0.0014018112560734153, + 0.0018113252008333802, + 3.277038922533393e-05, + -4.067507688887417e-05, + -0.0006657886551693082, + 0.0020185261964797974, + -0.0015718279173597693, + -0.0024743354879319668, + 0.0002824710390996188, + 0.003487984649837017, + 0.0005018783849664032, + -0.0022507067769765854, + -0.0005763396038673818, + 0.0007501582731492817, + 0.0011279437458142638, + 0.0008351888973265886, + 0.0006766966544091702, + 0.0011843631509691477, + -3.760406980291009e-05, + 0.003942208830267191, + 0.00019845661881845444, + 0.002232921775430441, + 0.004017248749732971, + 0.007412542589008808, + 0.0004585348069667816, + 0.007836693897843361, + 0.004988133907318115, + 0.002706740517169237, + 0.003983347676694393, + 0.006801050156354904, + 0.0037632600869983435, + 0.010236972942948341, + 0.00855157058686018, + 0.009376580826938152, + 0.006161993369460106, + 0.008103354834020138, + 0.011376159265637398, + 0.014972970820963383, + 0.012046518735587597, + 0.010956445708870888 + ], + "teaching_negative_gradient_cosine": [ + 0.0005255909636616707, + -0.0010781302116811275, + 0.0025069178082048893, + 0.0016919649206101894, + 0.00010788301005959511, + 2.1327403374016285e-05, + -0.000367272412404418, + 0.0003776442026719451, + 0.0016317082336172462, + 0.0014963049907237291, + -0.0005021097604185343, + -0.0002937753451988101, + -0.0004791772225871682, + 6.390659837052226e-05, + -0.001030154642648995, + 0.0014018112560734153, + 0.0018113252008333802, + 3.277038922533393e-05, + -4.067507688887417e-05, + -0.0006657886551693082, + 0.0020185261964797974, + -0.0015718279173597693, + -0.0024743354879319668, + 0.0002824710390996188, + 0.003487984649837017, + 0.0005018783849664032, + -0.0022507067769765854, + -0.0005763396038673818, + 0.0007501582731492817, + 0.0011279437458142638, + 0.0008351888973265886, + 0.0006766966544091702, + 0.0011843631509691477, + -3.760406980291009e-05, + 0.003942208830267191, + 0.00019845661881845444, + 0.002232921775430441, + 0.004017248749732971, + 0.007412542589008808, + 0.0004585348069667816, + 0.007836693897843361, + 0.004988133907318115, + 0.002706740517169237, + 0.003983347676694393, + 0.006801050156354904, + 0.0037632600869983435, + 0.010236972942948341, + 0.00855157058686018, + 0.009376580826938152, + 0.006161993369460106, + 0.008103354834020138, + 0.011376159265637398, + 0.014972970820963383, + 0.012046518735587597, + 0.010956445708870888 + ], + "wall_s": 0.20534896850585938 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 19.85203016052246 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 15.277210861816407 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 22.87974688232422 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 25.236587622680663 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 27.989678770751954 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 44.29404400390625 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 55.89321231445312 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 86.94409305175782 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 114.98562063476562 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 165.99049699707032 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 228.40711592285157 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 289.34972493164065 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 364.11808366210937 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 439.4125665820313 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 671.6069704296875 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 712.7547475976562 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 980.140315546875 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 1103.293306484375 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 1320.8525594921875 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 1465.6310321875 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 1796.9750183007814 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 1948.1184459765625 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 2287.01851765625 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 2444.085239296875 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 2908.4220490625 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 3158.44132859375 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 3313.92180625 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 3877.503080390625 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 4540.05339828125 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 4695.8629065625 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 5154.7014296875 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 5821.560145234375 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 6480.87629140625 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 6341.7844628125 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 6923.978560625 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 7701.088639375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 7469.7678303125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 8743.23543609375 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 8046.617881875 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 9276.7875521875 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 10553.15291 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 10569.499498125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 11961.35611875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 12400.59445 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 12074.1799596875 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 14217.12017125 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 13769.598210625 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 13573.48423125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 15667.26407375 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 16657.7972615625 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 15693.64605375 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 18511.232810625 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 18231.63009625 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 17852.489756875 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 18452.930683125 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 21131.5533175 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 21925.75662875 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 23009.8898825 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 22536.6942965625 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 23355.11246875 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 23856.826025 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 26181.8635275 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 25788.6849225 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 28495.095256875 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 28800.2621128125 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 29564.81689 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 28106.36562625 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 30395.82846625 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 34965.03667 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 32706.780568125 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 37163.961801875 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 37854.28123375 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 33942.487185 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 38028.11356375 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 39465.98959875 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 40007.19174 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 42236.5418225 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 42935.5670475 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 44710.2431825 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 46254.5428475 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 45069.667475 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 45259.70246625 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 45629.35346375 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 45828.0199975 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 51119.8721825 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 54685.955385 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 48313.62319375 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 53704.67348 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 49807.59677 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 54847.301465 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 52811.8402475 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 53685.5629375 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 58321.18193875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 53370.31019 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 55259.42770875 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 62743.202595 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 71849.919185 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 62031.9489825 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 68600.14554 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 68424.4290175 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 13175.8239165625 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 9052.00531375 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 8383.80109984375 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 8050.7802765625 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 7575.8674025 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 7245.3161803125 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 6778.7499434375 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 6560.2829946875 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 6089.2685134375 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 5857.6013596875 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 5703.1139596875 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 5420.600954375 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 5243.32939 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 5076.914825 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 5148.14045984375 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 5220.29229140625 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 4584.908725703125 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 5142.28931734375 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 5081.4689378125 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 4670.038472578125 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 4729.2374884375 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 4437.689307421875 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 4643.063570703125 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 4598.3138490625 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 4710.731543203125 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 4922.6924784375 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 4706.8065034375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 4620.84078453125 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 4983.67323234375 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 4512.0828665625 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 4907.38452609375 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 4981.26236390625 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 4804.82025484375 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 4912.26073234375 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 5047.9833865625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 4543.504403125 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 5049.753786875 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 4841.258673125 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 4753.69637796875 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 5182.0467875 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 5353.002407578125 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 4778.885925625 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 5801.38689421875 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 4708.11906 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 5412.79496265625 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 5151.65164765625 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 5323.728211171875 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 4849.5092225 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 5417.35654671875 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 4851.2205709375 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 1917.0817833203125 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 1419.874765390625 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 1290.1946449609375 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 1215.0555855859375 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 1163.6108630859376 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 1105.2684625390625 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 1058.7536784765625 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 1024.7534498828124 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 972.62020796875 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 953.9492185546875 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 929.1100657421875 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 911.06114390625 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 873.1208018945313 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 873.9455255078125 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 855.66897359375 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 832.5822851953125 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 820.8964893359375 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 809.5474932226563 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 785.56792296875 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 760.2091385546875 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 751.9481428515625 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 768.3446512304688 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 739.8024948632813 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 721.0074110742188 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 748.7323610351563 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 701.091491328125 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 703.9146436914062 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 720.5603126953125 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 660.6486718359375 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 693.0599039257812 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 679.6050608203125 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 692.9751318359375 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 683.6897475585937 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 668.1460813085937 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 667.1369373828124 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 662.4229653515625 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 699.9143864453125 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 630.793762109375 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 675.000643828125 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 607.7093490722656 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 658.6411365429688 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 657.6614402148438 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 666.3172574804687 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 655.2991161328125 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 624.0942075195312 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 620.8854083203125 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 639.8727234960937 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 638.6959756640625 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 617.4043270703125 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 615.680954453125 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.3488, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 502.805196875, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3339442688, + "peak_memory_reserved_bytes": 4676648960, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0348587036132812, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10378.02499294281, + "train_wall_s": 10375.962883234024 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 53248000000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2562961920000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json b/results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json new file mode 100644 index 0000000..828f83e --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json @@ -0,0 +1,2281 @@ +{ + "architecture": { + "adaptive_apical_parameters": 6389760, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 5324800 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0003577964930122511, + "innovation_negative_gradient_cosine": [ + 0.002081854734569788, + -6.581633351743221e-05, + 0.00041557435179129243, + 0.0008862123358994722, + -0.0005109517369419336, + 0.00042458175448700786, + 0.0027301099617034197, + 0.001137684565037489, + -0.0005082294228486717, + -0.0018658648477867246, + 0.0010939594358205795, + -0.0005554755916818976, + -0.0008758855401538312, + 0.001304553123190999, + -0.0017698571318760514, + 0.0007898603216744959, + 0.0013096316251903772, + 0.0004183952696621418, + -0.0008684020722284913, + -5.6430697441101074e-05, + -0.0011641886085271835, + -0.0017775511369109154, + 0.0027877171523869038, + 0.0006546652875840664, + 0.0018238320481032133, + 0.0007054642192088068, + -0.00016221619443967938, + 0.003270443296059966, + 0.0008029259042814374, + 0.0014108323957771063, + 0.002188350073993206, + 0.0003249559667892754, + 0.0020819089841097593, + 0.0025527561083436012, + 0.00035986059810966253, + 0.0027287080883979797, + -0.0007787201902829111, + 0.009754732251167297, + 0.005875065922737122, + 0.005507401190698147, + 0.00709145050495863, + 0.008006663993000984, + 0.011602742597460747, + 0.011238088831305504, + 0.011543793603777885, + 0.008512252941727638, + 0.014168035238981247, + 0.006046508438885212, + 0.00905943289399147, + 0.006326874252408743, + 0.00905679538846016, + 0.016640275716781616, + 0.019679304212331772, + 0.013295481912791729, + 0.017547961324453354 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.002081854734569788, + -6.581633351743221e-05, + 0.00041557435179129243, + 0.0008862123358994722, + -0.0005109517369419336, + 0.00042458175448700786, + 0.0027301099617034197, + 0.001137684565037489, + -0.0005082294228486717, + -0.0018658648477867246, + 0.0010939594358205795, + -0.0005554755916818976, + -0.0008758855401538312, + 0.001304553123190999, + -0.0017698571318760514, + 0.0007898603216744959, + 0.0013096316251903772, + 0.0004183952696621418, + -0.0008684020722284913, + -5.6430697441101074e-05, + -0.0011641886085271835, + -0.0017775511369109154, + 0.0027877171523869038, + 0.0006546652875840664, + 0.0018238320481032133, + 0.0007054642192088068, + -0.00016221619443967938, + 0.003270443296059966, + 0.0008029259042814374, + 0.0014108323957771063, + 0.002188350073993206, + 0.0003249559667892754, + 0.0020819089841097593, + 0.0025527561083436012, + 0.00035986059810966253, + 0.0027287080883979797, + -0.0007787201902829111, + 0.009754732251167297, + 0.005875065922737122, + 0.005507401190698147, + 0.00709145050495863, + 0.008006663993000984, + 0.011602742597460747, + 0.011238088831305504, + 0.011543793603777885, + 0.008512252941727638, + 0.014168035238981247, + 0.006046508438885212, + 0.00905943289399147, + 0.006326874252408743, + 0.00905679538846016, + 0.016640275716781616, + 0.019679304212331772, + 0.013295481912791729, + 0.017547961324453354 + ], + "teaching_negative_gradient_cosine": [ + 0.002081854734569788, + -6.581633351743221e-05, + 0.00041557435179129243, + 0.0008862123358994722, + -0.0005109517369419336, + 0.00042458175448700786, + 0.0027301099617034197, + 0.001137684565037489, + -0.0005082294228486717, + -0.0018658648477867246, + 0.0010939594358205795, + -0.0005554755916818976, + -0.0008758855401538312, + 0.001304553123190999, + -0.0017698571318760514, + 0.0007898603216744959, + 0.0013096316251903772, + 0.0004183952696621418, + -0.0008684020722284913, + -5.6430697441101074e-05, + -0.0011641886085271835, + -0.0017775511369109154, + 0.0027877171523869038, + 0.0006546652875840664, + 0.0018238320481032133, + 0.0007054642192088068, + -0.00016221619443967938, + 0.003270443296059966, + 0.0008029259042814374, + 0.0014108323957771063, + 0.002188350073993206, + 0.0003249559667892754, + 0.0020819089841097593, + 0.0025527561083436012, + 0.00035986059810966253, + 0.0027287080883979797, + -0.0007787201902829111, + 0.009754732251167297, + 0.005875065922737122, + 0.005507401190698147, + 0.00709145050495863, + 0.008006663993000984, + 0.011602742597460747, + 0.011238088831305504, + 0.011543793603777885, + 0.008512252941727638, + 0.014168035238981247, + 0.006046508438885212, + 0.00905943289399147, + 0.006326874252408743, + 0.00905679538846016, + 0.016640275716781616, + 0.019679304212331772, + 0.013295481912791729, + 0.017547961324453354 + ], + "wall_s": 0.2199254035949707 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 20.490372321777343 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 16.882818517456055 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 22.28293617919922 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 32.22360906982422 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 46.52625292114258 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 61.080078745117184 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 69.91611538208008 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 92.61200556030273 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 110.42092873291016 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 139.7512891064453 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 169.34016850585937 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 227.0122556982422 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 292.92110482421873 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 361.45023318359375 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 426.76469272460935 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 500.7480604199219 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 546.9457539648438 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 724.0486225195312 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 833.3311554785156 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 1041.0364372070312 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 1248.4833929296874 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 1350.1294877734374 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 1710.98063296875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 2209.1547815625 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 2441.934182890625 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 2529.25472796875 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 2887.71826140625 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 3126.71301359375 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 3453.0569534375 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 4313.30736875 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 5114.32865453125 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 5369.68033453125 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 5650.06040875 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 6635.95551671875 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 6293.6670246875 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 7230.133379375 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 8361.1046715625 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 8847.93686375 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 9081.9119225 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 10225.6603215625 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 10560.681274375 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 10014.091368125 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 12802.432016875 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 12514.8965871875 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 13962.953141875 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 15282.5079678125 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 14999.5047015625 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 15210.9302703125 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 17634.892895625 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 16799.24489625 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 19577.6702021875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 18435.589159375 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 22150.93953 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 19658.92122875 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 21042.8679015625 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 21857.45334375 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 25517.614735625 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 25244.549959375 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 26335.128986875 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 27385.4901175 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 28137.03504 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 27979.586675 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 33889.63128375 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 32506.89203625 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 34838.36573 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 32596.39202875 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 39124.6518253125 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 36587.51864875 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 38581.61796375 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 39131.21147625 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 39419.66956375 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 45149.3276175 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 43675.2919675 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 43209.59469 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 47960.806325 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 45578.00503375 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 50038.49809 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 50708.95643125 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 49882.63458 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 52694.5556475 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 52183.03786375 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 57154.52413375 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 61268.2211825 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 61062.56609 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 63559.830705 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 67166.77463 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 73203.6220075 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 62199.4245425 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 77680.6113875 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 64505.8588025 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 82246.832105 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 72240.0251025 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 74938.227655 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 73391.8108975 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 80380.57861 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 86301.0710025 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 80236.23262 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 86720.108125 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 85931.05704125 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 87120.586675 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 16651.018333125 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 12207.6655284375 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 11093.006951875 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 10483.7009175 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 10113.0312990625 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 9756.068164375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 9243.780345625 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 8713.2818921875 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 8324.015185 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 7787.7575653125 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 7442.3282865625 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 7383.6853378125 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 6795.67533390625 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 7012.378193125 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 6524.51093484375 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 6777.3454996875 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 6447.0069984375 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 5999.93104765625 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 6029.22547125 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 6563.4999215625 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 6040.0361803125 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 6728.580324375 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 6468.49931953125 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 6330.526211875 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 6213.6745078125 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 6347.12538546875 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 6139.44533984375 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 6085.03322859375 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 6121.663730625 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 6252.5325828125 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 6137.2040996875 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 5885.20040140625 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 6218.637215625 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 5887.972265 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 6469.8672965625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 5943.174265625 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 6216.42217015625 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 6384.10118359375 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 6455.990405 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 6119.31162203125 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 6515.5113884375 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 6712.253388125 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 6020.62121953125 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 6674.37236703125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 6140.965126953125 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 6570.85379765625 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 6375.9274646875 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 6504.94818796875 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 5907.9925953125 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 7468.96049375 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 2355.716592265625 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 1971.07486875 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 1849.134235625 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 1727.903703046875 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 1663.20844578125 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 1579.936839765625 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 1518.787508203125 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 1477.2234746484376 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 1400.1150776953125 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 1354.4668286328124 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 1321.101190546875 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 1299.769038359375 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 1246.1335008984374 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 1228.830918359375 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 1192.2152896679688 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 1191.7070167578124 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 1168.252881796875 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 1166.2085495507813 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 1154.250678828125 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 1107.4836101171875 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 1105.0455069921875 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 1100.120903515625 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 1066.6065143359374 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 1069.9054935546876 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 1059.8130787109376 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 1079.388514296875 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 1055.166579140625 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 1069.33503328125 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 1045.348050234375 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 1053.1074065429686 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 1058.7408469921875 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 1009.6501013671875 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 1018.4961114453125 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 1004.7097914453125 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 1027.2876579492188 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 1027.0266332617186 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 995.8798777734376 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 1015.6011597265625 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 972.1149674609375 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 977.882844765625 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 1008.7832816796875 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 974.944484921875 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 1007.7110688671875 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 969.6423557421875 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 980.5617981445313 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 966.952641484375 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 988.5392790625 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 1018.6668183007813 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 974.0764008007812 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 962.7301450390624 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.2603, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 1081.2588375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3339442688, + "peak_memory_reserved_bytes": 4676648960, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.035142660140991, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 10382.633543491364, + "train_wall_s": 10380.57179403305 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 53248000000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 2562961920000000 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json new file mode 100644 index 0000000..633f12f --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json @@ -0,0 +1,6861 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5010, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9995589315891266, + "feedback_forward_cosine": [ + 0.9997740387916565, + 0.9997309446334839, + 0.9997027516365051, + 0.9996247291564941, + 0.9997149705886841, + 0.9996957778930664, + 0.9997022151947021, + 0.9996531009674072, + 0.9997735023498535, + 0.9997079372406006, + 0.9997566342353821, + 0.9998109936714172, + 0.9996917247772217, + 0.9996817708015442, + 0.9997506141662598, + 0.9997068643569946, + 0.9997621774673462, + 0.9996962547302246, + 0.9998061060905457, + 0.9997254014015198, + 0.9997891187667847, + 0.9998679757118225, + 0.9998348355293274, + 0.9998195171356201, + 0.9998433589935303, + 0.9997998476028442, + 0.9998409748077393, + 0.9997413754463196, + 0.999759316444397, + 0.9994879961013794, + 0.9999843835830688 + ], + "feedback_forward_norm_ratio": [ + 0.9994954466819763, + 1.0000416040420532, + 0.998541533946991, + 1.0021997690200806, + 1.0001829862594604, + 1.0013530254364014, + 0.9990200400352478, + 1.0003763437271118, + 0.9990434050559998, + 1.0013920068740845, + 0.9990915060043335, + 1.0000687837600708, + 0.9990549683570862, + 1.0015116930007935, + 0.9992631673812866, + 1.0009570121765137, + 0.999589204788208, + 1.0007927417755127, + 0.9992879033088684, + 1.0010055303573608, + 0.9994683861732483, + 1.0001976490020752, + 0.9995712637901306, + 1.0007340908050537, + 0.9997577667236328, + 1.000835657119751, + 0.9999958872795105, + 1.0008939504623413, + 0.999650776386261, + 1.0013740062713623, + 1.0024473667144775 + ], + "feedback_forward_relative_error": [ + 0.021259063854813576, + 0.02319592982530594, + 0.02441164292395115, + 0.027515806257724762, + 0.023877011612057686, + 0.024722866714000702, + 0.02441096119582653, + 0.026350652799010277, + 0.02129196748137474, + 0.024228349328041077, + 0.022065987810492516, + 0.019444063305854797, + 0.024835975840687752, + 0.025296475738286972, + 0.022337041795253754, + 0.024245882406830788, + 0.021809294819831848, + 0.02467196434736252, + 0.019692709669470787, + 0.023471806198358536, + 0.020535696297883987, + 0.01625259779393673, + 0.01818172261118889, + 0.019018346443772316, + 0.0176985003054142, + 0.020035430788993835, + 0.01783774606883526, + 0.0227713230997324, + 0.02193632908165455, + 0.03205590322613716, + 0.006116573233157396 + ], + "innovation_negative_gradient_cosine": [ + 0.9995901584625244, + 0.9994624853134155, + 0.9996066093444824, + 0.9994260668754578, + 0.9996123313903809, + 0.9995615482330322, + 0.9996132850646973, + 0.9995354413986206, + 0.9996192455291748, + 0.99956214427948, + 0.9996346235275269, + 0.999651312828064, + 0.9996753334999084, + 0.9995397329330444, + 0.9996817111968994, + 0.9995582699775696, + 0.9996984004974365, + 0.9996146559715271, + 0.9997084736824036, + 0.9996262788772583, + 0.9997171759605408, + 0.9997403621673584, + 0.9997692108154297, + 0.9997376203536987, + 0.9997997879981995, + 0.9997826218605042, + 0.9998412132263184, + 0.9998517036437988, + 0.9999121427536011, + 0.9999598264694214, + 0.7429801821708679 + ], + "instruction_negative_gradient_cosine": [ + 0.9995901584625244, + 0.9994624853134155, + 0.9996066093444824, + 0.9994260668754578, + 0.9996123313903809, + 0.9995615482330322, + 0.9996132850646973, + 0.9995353817939758, + 0.9996192455291748, + 0.99956214427948, + 0.9996346235275269, + 0.999651312828064, + 0.9996753931045532, + 0.9995397329330444, + 0.9996817111968994, + 0.9995582699775696, + 0.9996983408927917, + 0.9996145963668823, + 0.9997084140777588, + 0.9996262788772583, + 0.9997172355651855, + 0.9997403621673584, + 0.9997692108154297, + 0.9997376203536987, + 0.9997997879981995, + 0.9997826218605042, + 0.9998412132263184, + 0.999851644039154, + 0.9999121427536011, + 0.9999598264694214, + 0.7659255862236023 + ], + "matched_negative_gradient_cosine": [ + 0.02436966821551323, + 0.023935535922646523, + 0.02684222161769867, + 0.04496397823095322, + 0.028166145086288452, + 0.012743436731398106, + 0.024146437644958496, + -0.004558200482279062, + 0.021830925717949867, + 0.03601580485701561, + 0.03358205407857895, + 0.003503045067191124, + 0.04407677799463272, + 0.06796478480100632, + 0.04375102370977402, + 0.08095117658376694, + 0.055207736790180206, + 0.15869669616222382, + 0.044433970004320145, + 0.18739309906959534, + 0.04294835776090622, + 0.0767122432589531, + 0.008538261987268925, + 0.25225597620010376, + 0.04945363849401474, + 0.29917144775390625, + 0.14587603509426117, + 0.28368568420410156, + 0.19282856583595276, + 0.24919679760932922, + 0.12054048478603363 + ], + "max_norm_match_direction_error": 2.384185791015625e-07, + "max_norm_match_relative_error": 1.67687119301263e-07, + "mean_feedback_forward_cosine": 0.9997495874281852, + "mean_feedback_forward_relative_error": 0.02198631041533043, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.08339470624923706, + "max_absolute_post_projection_soma_slope": 2.176913760365551e-08, + "max_absolute_pre_projection_soma_slope": 0.08339470624923706, + "max_positive_post_projection_soma_slope": 2.176913760365551e-08, + "min_post_projection_soma_slope": -9.30956645106562e-09, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4092123232960273e-08, + "pre_projection_traffic_rms_ratio": 0.007342756290308434 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.007342756290262722, + "raw_negative_gradient_cosine": [ + 0.02436966821551323, + 0.023935535922646523, + 0.02684222161769867, + 0.04496397823095322, + 0.028166145086288452, + 0.012743436731398106, + 0.024146437644958496, + -0.004558197688311338, + 0.021830927580595016, + 0.03601580485701561, + 0.03358205407857895, + 0.0035030432045459747, + 0.04407677799463272, + 0.06796477735042572, + 0.04375102370977402, + 0.08095117658376694, + 0.055207736790180206, + 0.15869669616222382, + 0.044433966279029846, + 0.18739309906959534, + 0.04294835776090622, + 0.0767122358083725, + 0.00853826105594635, + 0.25225594639778137, + 0.04945364594459534, + 0.29917144775390625, + 0.14587603509426117, + 0.28368568420410156, + 0.19282855093479156, + 0.24919679760932922, + 0.12054048478603363 + ], + "teaching_negative_gradient_cosine": [ + 0.9995901584625244, + 0.9994624853134155, + 0.9996066093444824, + 0.9994260668754578, + 0.9996123313903809, + 0.9995615482330322, + 0.9996132850646973, + 0.9995354413986206, + 0.9996192455291748, + 0.99956214427948, + 0.9996346235275269, + 0.999651312828064, + 0.9996753334999084, + 0.9995397329330444, + 0.9996817111968994, + 0.9995582699775696, + 0.9996984004974365, + 0.9996146559715271, + 0.9997084736824036, + 0.9996262788772583, + 0.9997171759605408, + 0.9997403621673584, + 0.9997692108154297, + 0.9997376203536987, + 0.9997997879981995, + 0.9997826218605042, + 0.9998412132263184, + 0.9998517036437988, + 0.9999121427536011, + 0.9999598264694214, + 0.7429801821708679 + ], + "traffic_instruction_rms_ratio": 22.42608294638183, + "used_negative_gradient_cosine": [ + 0.9995901584625244, + 0.9994624853134155, + 0.9996066093444824, + 0.9994260668754578, + 0.9996123313903809, + 0.9995615482330322, + 0.9996132850646973, + 0.9995354413986206, + 0.9996192455291748, + 0.99956214427948, + 0.9996346235275269, + 0.999651312828064, + 0.9996753334999084, + 0.9995397329330444, + 0.9996817111968994, + 0.9995582699775696, + 0.9996984004974365, + 0.9996146559715271, + 0.9997084736824036, + 0.9996262788772583, + 0.9997171759605408, + 0.9997403621673584, + 0.9997692108154297, + 0.9997376203536987, + 0.9997997879981995, + 0.9997826218605042, + 0.9998412132263184, + 0.9998517036437988, + 0.9999121427536011, + 0.9999598264694214, + 0.7429801821708679 + ], + "wall_s": 0.24200749397277832 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.355901837348938, + "mean_feedback_forward_cosine": 0.2852873273434178, + "mean_feedback_forward_relative_error": 1.1913391543972878, + "min_feedback_forward_cosine": 0.08586828410625458 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0013150885607349611, + "instruction_rms": 0.00131508856097854, + "raw_apical_rms": 0.03484235745422441, + "teaching_rms": 0.0013150885607349611, + "traffic_rms": 0.03478066642460851 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339476585388184, + "maximum_absolute_post_projection_soma_slope": 3.812822058080201e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.032019977820027e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0075063381139136446, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.0549319031143187 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.327340006828308, + "mean_feedback_forward_cosine": 0.34048358951845475, + "mean_feedback_forward_relative_error": 1.1447024422307168, + "min_feedback_forward_cosine": 0.12527257204055786 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0016963456007157778, + "instruction_rms": 0.0016963456004990569, + "raw_apical_rms": 0.04024585679608645, + "teaching_rms": 0.0016963456007157778, + "traffic_rms": 0.04020947866848 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.507070900354847e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7969390483677425e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0068364199827301, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.5279106087112426 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2859400510787964, + "mean_feedback_forward_cosine": 0.4163759865107075, + "mean_feedback_forward_relative_error": 1.0783473012908813, + "min_feedback_forward_cosine": 0.18110977113246918 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002580048269783251, + "instruction_rms": 0.0025800482699530124, + "raw_apical_rms": 0.04554578361803903, + "teaching_rms": 0.002580048269783251, + "traffic_rms": 0.04547182030609843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.415419851966362e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.754833210230822e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008881454455707012, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.2715791214370729 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2412482500076294, + "mean_feedback_forward_cosine": 0.4899324380582379, + "mean_feedback_forward_relative_error": 1.0095262863943655, + "min_feedback_forward_cosine": 0.23849225044250488 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034018786332155066, + "instruction_rms": 0.0034018786344267417, + "raw_apical_rms": 0.04921902618349008, + "teaching_rms": 0.0034018786332155066, + "traffic_rms": 0.04910040835566669 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.1035885800511096e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.717590752438206e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010448503530181886, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.0204214226150512 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.201474666595459, + "mean_feedback_forward_cosine": 0.5483782426003487, + "mean_feedback_forward_relative_error": 0.9508343819648989, + "min_feedback_forward_cosine": 0.29382187128067017 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003755928134117291, + "instruction_rms": 0.0037559281342010643, + "raw_apical_rms": 0.05244726618885307, + "teaching_rms": 0.003755928134117291, + "traffic_rms": 0.052311337931786044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.597585163106487e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6904462615328156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011300256942293823, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8440713566589355 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1691337823867798, + "mean_feedback_forward_cosine": 0.5943051163227328, + "mean_feedback_forward_relative_error": 0.9018531864689242, + "min_feedback_forward_cosine": 0.3400437831878662 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038564600137755358, + "instruction_rms": 0.003856460012589805, + "raw_apical_rms": 0.05468812043183341, + "teaching_rms": 0.0038564600137755358, + "traffic_rms": 0.054550762737266244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.883540682953935e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6751136142438453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011332458019579068, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7406229090690613 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1405353546142578, + "mean_feedback_forward_cosine": 0.631730503612949, + "mean_feedback_forward_relative_error": 0.8595834172541096, + "min_feedback_forward_cosine": 0.3737984299659729 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003907344717177833, + "instruction_rms": 0.0039073447173506816, + "raw_apical_rms": 0.056909366153420994, + "teaching_rms": 0.003907344717177833, + "traffic_rms": 0.05677376059815443 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.706522105379918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6758073235919733e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011085048338354104, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6593398728179931 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1113486289978027, + "mean_feedback_forward_cosine": 0.6641632635747233, + "mean_feedback_forward_relative_error": 0.8209669763042081, + "min_feedback_forward_cosine": 0.4066813588142395 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003990091048657211, + "instruction_rms": 0.003990091048852472, + "raw_apical_rms": 0.05804253738402756, + "teaching_rms": 0.003990091048657211, + "traffic_rms": 0.05790395606961878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.4732874354403975e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.669251196960308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009877755518766499, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6113407627677917 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.082719087600708, + "mean_feedback_forward_cosine": 0.6921723225424367, + "mean_feedback_forward_relative_error": 0.7858609268742223, + "min_feedback_forward_cosine": 0.43843305110931396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003964494074421636, + "instruction_rms": 0.003964494074318108, + "raw_apical_rms": 0.05932042115448445, + "teaching_rms": 0.003964494074421636, + "traffic_rms": 0.05918637979609914 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.906145806809036e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6668357684750913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01073259391523067, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5666326992416382 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0535595417022705, + "mean_feedback_forward_cosine": 0.7170614888591151, + "mean_feedback_forward_relative_error": 0.7533362844297963, + "min_feedback_forward_cosine": 0.46895086765289307 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040015759836207945, + "instruction_rms": 0.004001575984398696, + "raw_apical_rms": 0.06059908415036365, + "teaching_rms": 0.0040015759836207945, + "traffic_rms": 0.06046504577138352 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.141637433008327e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6525504298387055e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011589405876446466, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5297036050415039 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0252355337142944, + "mean_feedback_forward_cosine": 0.7390876143209396, + "mean_feedback_forward_relative_error": 0.7232973796706046, + "min_feedback_forward_cosine": 0.497922420501709 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004010087298212577, + "instruction_rms": 0.004010087298705321, + "raw_apical_rms": 0.06164359882323319, + "teaching_rms": 0.004010087298212577, + "traffic_rms": 0.0615113754899271 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.4137283222435144e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6440752975873324e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01090744082179042, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5002654731845856 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9966118931770325, + "mean_feedback_forward_cosine": 0.7589489529209752, + "mean_feedback_forward_relative_error": 0.6948745740998176, + "min_feedback_forward_cosine": 0.5261656045913696 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004010168196352792, + "instruction_rms": 0.004010168195776317, + "raw_apical_rms": 0.062487694327737764, + "teaching_rms": 0.004010168196352792, + "traffic_rms": 0.06235728861555303 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.349732935475913e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6399026551057674e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010842519805162293, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.47606026063919066 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9682766199111938, + "mean_feedback_forward_cosine": 0.778004611692121, + "mean_feedback_forward_relative_error": 0.6664577011139162, + "min_feedback_forward_cosine": 0.5530743598937988 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040969481889085516, + "instruction_rms": 0.004096948188787894, + "raw_apical_rms": 0.0628274022822907, + "teaching_rms": 0.0040969481889085516, + "traffic_rms": 0.06269190935063933 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9702793113983716e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6359904081027137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010646548982429315, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4559270332813263 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9407852292060852, + "mean_feedback_forward_cosine": 0.7948982311833289, + "mean_feedback_forward_relative_error": 0.6403621579370191, + "min_feedback_forward_cosine": 0.5782277584075928 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004063829557878184, + "instruction_rms": 0.004063829558396398, + "raw_apical_rms": 0.06372131699376611, + "teaching_rms": 0.004063829557878184, + "traffic_rms": 0.06358992870339157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0118812333057576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6278581302425092e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010365332279405429, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.43522120515823365 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9140377044677734, + "mean_feedback_forward_cosine": 0.8103813913560682, + "mean_feedback_forward_relative_error": 0.6153127503010535, + "min_feedback_forward_cosine": 0.6020639538764954 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004040252380590241, + "instruction_rms": 0.00404025238101508, + "raw_apical_rms": 0.06474712431372905, + "teaching_rms": 0.004040252380590241, + "traffic_rms": 0.06461911420571795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8675136931610723e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.627431558462012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010635548573964435, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.41694536555290224 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8878570199012756, + "mean_feedback_forward_cosine": 0.8244789319653665, + "mean_feedback_forward_relative_error": 0.591701201373531, + "min_feedback_forward_cosine": 0.6244533658027649 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040404580250400085, + "instruction_rms": 0.004040458025074786, + "raw_apical_rms": 0.06552216446823896, + "teaching_rms": 0.0040404580250400085, + "traffic_rms": 0.06539567100592737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0631628789024035e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.627446215788966e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010049432767590133, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.40558433619499207 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8616690635681152, + "mean_feedback_forward_cosine": 0.8373137885524381, + "mean_feedback_forward_relative_error": 0.5692409392326109, + "min_feedback_forward_cosine": 0.6465263366699219 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040150428332068715, + "instruction_rms": 0.004015042833261711, + "raw_apical_rms": 0.06618050256427714, + "teaching_rms": 0.0040150428332068715, + "traffic_rms": 0.06605664098845627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.780158325776938e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6207527320206334e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009688953653773081, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.38887233457565307 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8360998034477234, + "mean_feedback_forward_cosine": 0.8491213552413448, + "mean_feedback_forward_relative_error": 0.5479545554807109, + "min_feedback_forward_cosine": 0.6672181487083435 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004026569946400506, + "instruction_rms": 0.004026569946748301, + "raw_apical_rms": 0.06651230574523131, + "teaching_rms": 0.004026569946400506, + "traffic_rms": 0.06638833433219032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.406674859400937e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6266179546135666e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010354216357478097, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3751208915233612 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8097883462905884, + "mean_feedback_forward_cosine": 0.8604870361666526, + "mean_feedback_forward_relative_error": 0.5265913466292043, + "min_feedback_forward_cosine": 0.6878430247306824 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004045563500797783, + "instruction_rms": 0.0040455635008758584, + "raw_apical_rms": 0.06692520747789484, + "teaching_rms": 0.004045563500797783, + "traffic_rms": 0.06680078995742675 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.9425661907112044e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6226443160509122e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009212847204157338, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.36499167021751405 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7832959294319153, + "mean_feedback_forward_cosine": 0.8708358137838302, + "mean_feedback_forward_relative_error": 0.5062668155277928, + "min_feedback_forward_cosine": 0.7075591683387756 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004016434326381081, + "instruction_rms": 0.004016434326946165, + "raw_apical_rms": 0.06744402887766679, + "teaching_rms": 0.004016434326381081, + "traffic_rms": 0.06732237993706922 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6816575626753547e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6178786900768774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010256992207468606, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.358980114402771 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7576461434364319, + "mean_feedback_forward_cosine": 0.8800988158872051, + "mean_feedback_forward_relative_error": 0.48746528788920374, + "min_feedback_forward_cosine": 0.7260517477989197 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0039145930922363245, + "instruction_rms": 0.003914593089588133, + "raw_apical_rms": 0.06775718054868864, + "teaching_rms": 0.0039145930922363245, + "traffic_rms": 0.06764206603663708 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9244217714108345e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.614565286792486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01007597096411203, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.34363371705055235 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7333029508590698, + "mean_feedback_forward_cosine": 0.8887851468978389, + "mean_feedback_forward_relative_error": 0.46922569265288694, + "min_feedback_forward_cosine": 0.7430939674377441 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003968343398757923, + "instruction_rms": 0.003968343399322759, + "raw_apical_rms": 0.06766371910018273, + "teaching_rms": 0.003968343398757923, + "traffic_rms": 0.06754525459794881 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8113346317582e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6172171000390134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010355066739213608, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3329271225643158 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.708159327507019, + "mean_feedback_forward_cosine": 0.8971072089287543, + "mean_feedback_forward_relative_error": 0.4510384975902496, + "min_feedback_forward_cosine": 0.7603352069854736 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0039925677956464765, + "instruction_rms": 0.003992567794789574, + "raw_apical_rms": 0.06811845471550859, + "teaching_rms": 0.0039925677956464765, + "traffic_rms": 0.06799902373379944 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.876948279606495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6183721195731257e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010109405315751363, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3241496332168579 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6836541295051575, + "mean_feedback_forward_cosine": 0.9048887452771587, + "mean_feedback_forward_relative_error": 0.43351450129862756, + "min_feedback_forward_cosine": 0.7764149904251099 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003993437418829618, + "instruction_rms": 0.003993437417674439, + "raw_apical_rms": 0.06883859664801428, + "teaching_rms": 0.003993437418829618, + "traffic_rms": 0.06872035719581852 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9346761465376403e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6083075734488722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010582388246811042, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.31755477598190307 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6599841713905334, + "mean_feedback_forward_cosine": 0.9118670552007614, + "mean_feedback_forward_relative_error": 0.4171058021245464, + "min_feedback_forward_cosine": 0.7915564775466919 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003945102122322032, + "instruction_rms": 0.003945102120359529, + "raw_apical_rms": 0.0692187785577354, + "teaching_rms": 0.003945102122322032, + "traffic_rms": 0.06910412269078686 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0883885671073585e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.605825593003453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010479713474136175, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31450688428878787 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6355026960372925, + "mean_feedback_forward_cosine": 0.9185291605611001, + "mean_feedback_forward_relative_error": 0.40078419110467356, + "min_feedback_forward_cosine": 0.8066827654838562 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003986443019869842, + "instruction_rms": 0.003986443019222523, + "raw_apical_rms": 0.06964720194991254, + "teaching_rms": 0.003986443019869842, + "traffic_rms": 0.06953091354488793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8504461013767468e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6060907137306867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010706610008114553, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31076201431274414 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6138848662376404, + "mean_feedback_forward_cosine": 0.9244122370596854, + "mean_feedback_forward_relative_error": 0.3858543773812632, + "min_feedback_forward_cosine": 0.8195127844810486 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00395727869577107, + "instruction_rms": 0.003957278694457989, + "raw_apical_rms": 0.07003219610683441, + "teaching_rms": 0.00395727869577107, + "traffic_rms": 0.06991807275540245 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.190206854242206e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.609355681670642e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010017265818399526, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2911986347579956 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5914310812950134, + "mean_feedback_forward_cosine": 0.9300016299370797, + "mean_feedback_forward_relative_error": 0.3711627169001487, + "min_feedback_forward_cosine": 0.832229495048523 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003989525615953841, + "instruction_rms": 0.003989525614823946, + "raw_apical_rms": 0.07037566354420195, + "teaching_rms": 0.003989525615953841, + "traffic_rms": 0.0702602044086338 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.057853703580804e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.611923204240864e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010354357278439079, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2919534160041809 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5701101422309875, + "mean_feedback_forward_cosine": 0.9351586353394293, + "mean_feedback_forward_relative_error": 0.3570136045736651, + "min_feedback_forward_cosine": 0.8438876867294312 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003916522260135462, + "instruction_rms": 0.0039165222608802015, + "raw_apical_rms": 0.07069943514489989, + "teaching_rms": 0.003916522260135462, + "traffic_rms": 0.07058853849817812 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.2089534585111323e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6070630583397585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009864038319295616, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2853457137966156 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5485604405403137, + "mean_feedback_forward_cosine": 0.939991093450977, + "mean_feedback_forward_relative_error": 0.343268955907514, + "min_feedback_forward_cosine": 0.8552882671356201 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038888734481912733, + "instruction_rms": 0.0038888734484806642, + "raw_apical_rms": 0.07119314528072004, + "teaching_rms": 0.0038888734481912733, + "traffic_rms": 0.07108478376763376 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.1641795910909423e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6068949907262807e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009344873604752581, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2808817547035217 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5274695754051208, + "mean_feedback_forward_cosine": 0.9444268884197358, + "mean_feedback_forward_relative_error": 0.3302037600548037, + "min_feedback_forward_cosine": 0.8659871816635132 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038612350577803705, + "instruction_rms": 0.003861235060363108, + "raw_apical_rms": 0.07156784535807284, + "teaching_rms": 0.0038612350577803705, + "traffic_rms": 0.07146112307308779 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9213742536171594e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.599961771195114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00899707845930508, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27626795665740966 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5078015923500061, + "mean_feedback_forward_cosine": 0.9485022002650846, + "mean_feedback_forward_relative_error": 0.31772861389383195, + "min_feedback_forward_cosine": 0.8757166862487793 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038673789960190917, + "instruction_rms": 0.0038673789958513626, + "raw_apical_rms": 0.07154200206454321, + "teaching_rms": 0.0038673789960190917, + "traffic_rms": 0.07143512845786441 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.8276232910684485e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.599828349729547e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009012329077991727, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2691772640275955 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4882529675960541, + "mean_feedback_forward_cosine": 0.9524526557614726, + "mean_feedback_forward_relative_error": 0.30515540727684576, + "min_feedback_forward_cosine": 0.8849746584892273 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004001714659556043, + "instruction_rms": 0.004001714660291499, + "raw_apical_rms": 0.07188466507873913, + "teaching_rms": 0.004001714659556043, + "traffic_rms": 0.07177060212648059 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6061382385478282e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6010016624015026e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008843690193080572, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.26975091490745545 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.47038230299949646, + "mean_feedback_forward_cosine": 0.9559128226772431, + "mean_feedback_forward_relative_error": 0.29372032347225374, + "min_feedback_forward_cosine": 0.8931869268417358 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003866794331814295, + "instruction_rms": 0.0038667943315269146, + "raw_apical_rms": 0.07248943034839585, + "teaching_rms": 0.003866794331814295, + "traffic_rms": 0.07238389863981391 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.1972885733134717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5987922491578846e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008119160435629602, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2609949604797363 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4523027837276459, + "mean_feedback_forward_cosine": 0.9591905647708524, + "mean_feedback_forward_relative_error": 0.2825319216616692, + "min_feedback_forward_cosine": 0.9011703133583069 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003888217473177897, + "instruction_rms": 0.003888217472393469, + "raw_apical_rms": 0.07271104291605943, + "teaching_rms": 0.003888217473177897, + "traffic_rms": 0.07260443796315413 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.970719492623175e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5943847238346735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008716405861196121, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2555421365642548 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43530547618865967, + "mean_feedback_forward_cosine": 0.962148618313574, + "mean_feedback_forward_relative_error": 0.27193370629702845, + "min_feedback_forward_cosine": 0.9083752632141113 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038472607075104286, + "instruction_rms": 0.003847260708558824, + "raw_apical_rms": 0.07311550607523036, + "teaching_rms": 0.0038472607075104286, + "traffic_rms": 0.07301197379605809 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.564729939535937e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.601760873368359e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008733120888548558, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2510679250240326 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41786566376686096, + "mean_feedback_forward_cosine": 0.9650170514660497, + "mean_feedback_forward_relative_error": 0.26134377453596364, + "min_feedback_forward_cosine": 0.9154398441314697 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00391465809543543, + "instruction_rms": 0.003914658094983083, + "raw_apical_rms": 0.0729227545519694, + "teaching_rms": 0.00391465809543543, + "traffic_rms": 0.07281504936157407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.437560636143644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5986736342805637e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008870490007415485, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2534460880565643 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4016924798488617, + "mean_feedback_forward_cosine": 0.967648027404662, + "mean_feedback_forward_relative_error": 0.2512254726982886, + "min_feedback_forward_cosine": 0.9217399954795837 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038684681372659377, + "instruction_rms": 0.0038684681367641954, + "raw_apical_rms": 0.07289484821777953, + "teaching_rms": 0.0038684681372659377, + "traffic_rms": 0.07278930519585007 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9372056786769463e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.599444901215317e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009470029783169135, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24605752365112304 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38644707202911377, + "mean_feedback_forward_cosine": 0.9700227052934708, + "mean_feedback_forward_relative_error": 0.2417544020760444, + "min_feedback_forward_cosine": 0.9275497198104858 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038557534338615443, + "instruction_rms": 0.0038557534328940925, + "raw_apical_rms": 0.0730733501443275, + "teaching_rms": 0.0038557534338615443, + "traffic_rms": 0.07296899002912055 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.2965522816775774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6022807170896294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009324111192015686, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.24045695706367493 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37039729952812195, + "mean_feedback_forward_cosine": 0.972320762372786, + "mean_feedback_forward_relative_error": 0.2322382777929306, + "min_feedback_forward_cosine": 0.9333997964859009 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003863954479881888, + "instruction_rms": 0.003863954479368754, + "raw_apical_rms": 0.07327049786244774, + "teaching_rms": 0.003863954479881888, + "traffic_rms": 0.07316580763237819 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8939748375478302e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.593296119595771e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008987748031136278, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24438991549015046 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3552638292312622, + "mean_feedback_forward_cosine": 0.9744124393309316, + "mean_feedback_forward_relative_error": 0.22328707047047153, + "min_feedback_forward_cosine": 0.9386234879493713 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003845642755001636, + "instruction_rms": 0.003845642753481902, + "raw_apical_rms": 0.07362757664145705, + "teaching_rms": 0.003845642755001636, + "traffic_rms": 0.07352450126421374 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.74471716465996e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6047489986458367e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008401534127999766, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2353876228237152 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34149542450904846, + "mean_feedback_forward_cosine": 0.9762915392075816, + "mean_feedback_forward_relative_error": 0.21486841190245845, + "min_feedback_forward_cosine": 0.9432387351989746 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038190742704273363, + "instruction_rms": 0.0038190742694001827, + "raw_apical_rms": 0.07370958562685571, + "teaching_rms": 0.0038190742704273363, + "traffic_rms": 0.07360805408494495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5660757074774665e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.59289747540714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008132971699278936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23211295499801635 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3280247747898102, + "mean_feedback_forward_cosine": 0.9780537947531669, + "mean_feedback_forward_relative_error": 0.206693280368082, + "min_feedback_forward_cosine": 0.9475867748260498 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003739299371577429, + "instruction_rms": 0.003739299370138564, + "raw_apical_rms": 0.07458029297740579, + "teaching_rms": 0.003739299371577429, + "traffic_rms": 0.07448380664695446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.3749898875612416e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.593370065675669e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008140765757782584, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2319326350736618 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31559228897094727, + "mean_feedback_forward_cosine": 0.9796506147230825, + "mean_feedback_forward_relative_error": 0.19898946174690801, + "min_feedback_forward_cosine": 0.9514440298080444 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038072827260089244, + "instruction_rms": 0.003807282726619384, + "raw_apical_rms": 0.07482573855703395, + "teaching_rms": 0.0038072827260089244, + "traffic_rms": 0.07472613621102088 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3296335527334122e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5941347562912665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008329016918594521, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22890997206687927 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3033545911312103, + "mean_feedback_forward_cosine": 0.9811681547472554, + "mean_feedback_forward_relative_error": 0.19137271349468538, + "min_feedback_forward_cosine": 0.9550760388374329 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037958648688829685, + "instruction_rms": 0.003795864868612557, + "raw_apical_rms": 0.07477883365826042, + "teaching_rms": 0.0037958648688829685, + "traffic_rms": 0.07467966679215148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.808553212219067e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589216325241839e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007714240284789253, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.22528609904766084 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2914389371871948, + "mean_feedback_forward_cosine": 0.9825650607385943, + "mean_feedback_forward_relative_error": 0.18408100931875168, + "min_feedback_forward_cosine": 0.958500862121582 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037406374919987852, + "instruction_rms": 0.003740637492317074, + "raw_apical_rms": 0.07476868493166478, + "teaching_rms": 0.0037406374919987852, + "traffic_rms": 0.0746723477867823 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.6830321903380536e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591689095207291e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008549012311788154, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.22341396438598632 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28000912070274353, + "mean_feedback_forward_cosine": 0.9838960247655069, + "mean_feedback_forward_relative_error": 0.17688635784772136, + "min_feedback_forward_cosine": 0.9616625905036926 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003788429335074308, + "instruction_rms": 0.003788429334407076, + "raw_apical_rms": 0.07482689082631115, + "teaching_rms": 0.003788429335074308, + "traffic_rms": 0.07472843505891101 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9951458202504e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5937123684712816e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007491225114115658, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.22198218373775483 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26914381980895996, + "mean_feedback_forward_cosine": 0.9850757429676671, + "mean_feedback_forward_relative_error": 0.17024329340746325, + "min_feedback_forward_cosine": 0.9645553827285767 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037051435019831666, + "instruction_rms": 0.003705143501034956, + "raw_apical_rms": 0.07531095649511314, + "teaching_rms": 0.0037051435019831666, + "traffic_rms": 0.07521728873562585 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.489898953683678e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587536353653519e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007733585194618078, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21495834109783174 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25782403349876404, + "mean_feedback_forward_cosine": 0.9862037743291547, + "mean_feedback_forward_relative_error": 0.16367973831872787, + "min_feedback_forward_cosine": 0.9674678444862366 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037595492592871547, + "instruction_rms": 0.0037595492592439827, + "raw_apical_rms": 0.07542204334045838, + "teaching_rms": 0.0037595492592871547, + "traffic_rms": 0.07532539366328513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.002826787223967e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5870614450875012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007583728806009881, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21523304907798768 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24763494729995728, + "mean_feedback_forward_cosine": 0.9872592553015678, + "mean_feedback_forward_relative_error": 0.15727029548537347, + "min_feedback_forward_cosine": 0.9699530601501465 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037993633421569244, + "instruction_rms": 0.003799363343137188, + "raw_apical_rms": 0.07581127332624338, + "teaching_rms": 0.0037993633421569244, + "traffic_rms": 0.07571363311250577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.2597214766051366e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.590768476185006e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00738441074489908, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21348953110694885 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2375047653913498, + "mean_feedback_forward_cosine": 0.9882402016270545, + "mean_feedback_forward_relative_error": 0.15109317533431516, + "min_feedback_forward_cosine": 0.9723342657089233 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003758042514036572, + "instruction_rms": 0.0037580425138219525, + "raw_apical_rms": 0.07634999026301606, + "teaching_rms": 0.003758042514036572, + "traffic_rms": 0.07625514573251428 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.415420752299724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5836734822757476e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00748790960677364, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.21254394674777985 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22811542451381683, + "mean_feedback_forward_cosine": 0.9891276974831859, + "mean_feedback_forward_relative_error": 0.14526261629596834, + "min_feedback_forward_cosine": 0.9744594097137451 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037472381875839783, + "instruction_rms": 0.0037472381874510312, + "raw_apical_rms": 0.07646375429890218, + "teaching_rms": 0.0037472381875839783, + "traffic_rms": 0.0763691150155151 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.432652301820326e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5846385280689522e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007542917084654997, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.21006733677864076 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21926471590995789, + "mean_feedback_forward_cosine": 0.9899317391457096, + "mean_feedback_forward_relative_error": 0.13976371672845656, + "min_feedback_forward_cosine": 0.9763983488082886 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003704912956609882, + "instruction_rms": 0.003704912956245514, + "raw_apical_rms": 0.07651026704923802, + "teaching_rms": 0.003704912956609882, + "traffic_rms": 0.07641752630726097 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.703242512642646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.584566921971642e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008138022357159809, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20507358961105346 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21055907011032104, + "mean_feedback_forward_cosine": 0.9906903543779927, + "mean_feedback_forward_relative_error": 0.1343689361406911, + "min_feedback_forward_cosine": 0.9782177209854126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037610988106757635, + "instruction_rms": 0.0037610988121124355, + "raw_apical_rms": 0.07677379383636217, + "teaching_rms": 0.0037610988106757635, + "traffic_rms": 0.07667893970489342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.442420393184875e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5803364830292135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007625236560505473, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.20681451694488526 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20280465483665466, + "mean_feedback_forward_cosine": 0.9913615872783046, + "mean_feedback_forward_relative_error": 0.1294098575749705, + "min_feedback_forward_cosine": 0.9797812104225159 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036806931411789245, + "instruction_rms": 0.003680693140367778, + "raw_apical_rms": 0.07706194060510733, + "teaching_rms": 0.0036806931411789245, + "traffic_rms": 0.07697109683093753 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7282942127726528e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5784764462732706e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007894901263081608, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19486241562843323 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19470247626304626, + "mean_feedback_forward_cosine": 0.9920127584088233, + "mean_feedback_forward_relative_error": 0.12443118650586374, + "min_feedback_forward_cosine": 0.9813491702079773 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037299922851433396, + "instruction_rms": 0.003729992284147289, + "raw_apical_rms": 0.0772884664793332, + "teaching_rms": 0.0037299922851433396, + "traffic_rms": 0.07719614313523333 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.317347818665439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.576550471129702e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007761257460099643, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.2058708153629303 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18697470426559448, + "mean_feedback_forward_cosine": 0.9926183665952375, + "mean_feedback_forward_relative_error": 0.11962047315412952, + "min_feedback_forward_cosine": 0.9827912449836731 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003750574471397808, + "instruction_rms": 0.00375057447042929, + "raw_apical_rms": 0.0772948126515323, + "teaching_rms": 0.003750574471397808, + "traffic_rms": 0.07720106588352636 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5107691925541076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5747330342699073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00784233930591821, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20121257962226868 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17957532405853271, + "mean_feedback_forward_cosine": 0.9931631722757893, + "mean_feedback_forward_relative_error": 0.11511752201664832, + "min_feedback_forward_cosine": 0.9841135740280151 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003662668878217999, + "instruction_rms": 0.0036626688795714023, + "raw_apical_rms": 0.07791188188895462, + "teaching_rms": 0.003662668878217999, + "traffic_rms": 0.07782319173571092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.738331161822316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5821882872903312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008064510568461034, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.194944077501297 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1722871959209442, + "mean_feedback_forward_cosine": 0.9936728919706037, + "mean_feedback_forward_relative_error": 0.11074392161061687, + "min_feedback_forward_cosine": 0.9853686094284058 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003690065660985786, + "instruction_rms": 0.003690065661159514, + "raw_apical_rms": 0.07814971609262164, + "teaching_rms": 0.003690065660985786, + "traffic_rms": 0.07806017785163784 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.187626518297293e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5703452872848636e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00826419011511355, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19906718844413757 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16538487374782562, + "mean_feedback_forward_cosine": 0.994152388265056, + "mean_feedback_forward_relative_error": 0.10644797512119816, + "min_feedback_forward_cosine": 0.9865068793296814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037220134810003994, + "instruction_rms": 0.0037220134805156894, + "raw_apical_rms": 0.07861880500200948, + "teaching_rms": 0.0037220134810003994, + "traffic_rms": 0.07852735762761377 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8621364833725238e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.572861628952369e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00804488518376896, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19197324133872987 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15865659713745117, + "mean_feedback_forward_cosine": 0.9945860466649455, + "mean_feedback_forward_relative_error": 0.10243177978742507, + "min_feedback_forward_cosine": 0.9875701665878296 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036756967838200588, + "instruction_rms": 0.0036756967835092067, + "raw_apical_rms": 0.07852890397506984, + "teaching_rms": 0.0036756967838200588, + "traffic_rms": 0.0784398471730066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.663713516426469e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5753745646322497e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008130088370309919, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18818937430143357 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15215329825878143, + "mean_feedback_forward_cosine": 0.9949974379231853, + "mean_feedback_forward_relative_error": 0.09845988680758784, + "min_feedback_forward_cosine": 0.9885643720626831 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036796930262741555, + "instruction_rms": 0.0036796930261788255, + "raw_apical_rms": 0.07875844023162022, + "teaching_rms": 0.0036796930262741555, + "traffic_rms": 0.07866972363818663 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.487852850980744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5738110710633882e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008032919081584065, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.1944829085326195 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14598067104816437, + "mean_feedback_forward_cosine": 0.995369057501516, + "mean_feedback_forward_relative_error": 0.09472833105152653, + "min_feedback_forward_cosine": 0.9894676208496094 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036404975762606376, + "instruction_rms": 0.003640497575796741, + "raw_apical_rms": 0.07908420325577573, + "teaching_rms": 0.0036404975762606376, + "traffic_rms": 0.0789976864468624 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.952219801954925e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5681745764662834e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008162281633851699, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1887198252916336 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14030863344669342, + "mean_feedback_forward_cosine": 0.9957132031840663, + "mean_feedback_forward_relative_error": 0.09113867484754132, + "min_feedback_forward_cosine": 0.9902660250663757 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003696651065230788, + "instruction_rms": 0.0036966510648144247, + "raw_apical_rms": 0.07930500808926666, + "teaching_rms": 0.003696651065230788, + "traffic_rms": 0.07921613079011502 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.631678697184725e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5692986619211024e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008451304282934901, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18966873677253723 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13458269834518433, + "mean_feedback_forward_cosine": 0.9960341722734513, + "mean_feedback_forward_relative_error": 0.08764962161019925, + "min_feedback_forward_cosine": 0.9910438656806946 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037063250873448266, + "instruction_rms": 0.003706325087279222, + "raw_apical_rms": 0.07930357355568904, + "teaching_rms": 0.0037063250873448266, + "traffic_rms": 0.07921465110548796 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.706752155745562e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.573267149462498e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008419355987524054, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18653674884796143 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1290910542011261, + "mean_feedback_forward_cosine": 0.996338386689463, + "mean_feedback_forward_relative_error": 0.08421859788077493, + "min_feedback_forward_cosine": 0.9917571544647217 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003755409094338965, + "instruction_rms": 0.003755409094812695, + "raw_apical_rms": 0.07940073871020314, + "teaching_rms": 0.003755409094338965, + "traffic_rms": 0.07930913514804784 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.05380716270065e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5659793860099052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008260306812628704, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18692917158126832 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12408643960952759, + "mean_feedback_forward_cosine": 0.9966049117426719, + "mean_feedback_forward_relative_error": 0.08109183909912263, + "min_feedback_forward_cosine": 0.992379903793335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036829986264340555, + "instruction_rms": 0.0036829986266643145, + "raw_apical_rms": 0.07962739066348803, + "teaching_rms": 0.0036829986264340555, + "traffic_rms": 0.07953967541359108 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0581485788161444e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.566883061778126e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00837261943244417, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1866636373615265 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11915253847837448, + "mean_feedback_forward_cosine": 0.9968554089146275, + "mean_feedback_forward_relative_error": 0.07803975189885785, + "min_feedback_forward_cosine": 0.9929676055908203 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036707826855637412, + "instruction_rms": 0.0036707826862155324, + "raw_apical_rms": 0.0796717360471924, + "teaching_rms": 0.0036707826855637412, + "traffic_rms": 0.07958410317724517 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.031730244629216e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5744897587154944e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00838858586876981, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18337094452857972 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11427207291126251, + "mean_feedback_forward_cosine": 0.9971046255480859, + "mean_feedback_forward_relative_error": 0.07488689048876686, + "min_feedback_forward_cosine": 0.9935302138328552 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037297102686179293, + "instruction_rms": 0.003729710267922763, + "raw_apical_rms": 0.0799029147860724, + "teaching_rms": 0.0037297102686179293, + "traffic_rms": 0.07981305009970575 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 4.280405718759539e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5752729057833483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008417601092586454, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18663185458660125 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10964248329401016, + "mean_feedback_forward_cosine": 0.9973233265261496, + "mean_feedback_forward_relative_error": 0.07200624475315694, + "min_feedback_forward_cosine": 0.9940427541732788 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036650287420353206, + "instruction_rms": 0.003665028741187252, + "raw_apical_rms": 0.07991056399672092, + "teaching_rms": 0.0036650287420353206, + "traffic_rms": 0.07982378474357117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.784740438244171e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5773465020243927e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00826227411471984, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.18453536989212035 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1053331270813942, + "mean_feedback_forward_cosine": 0.9975196257714303, + "mean_feedback_forward_relative_error": 0.06931946222339908, + "min_feedback_forward_cosine": 0.9945000410079956 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003597899004171556, + "instruction_rms": 0.0035978990040468136, + "raw_apical_rms": 0.08001944314149088, + "teaching_rms": 0.003597899004171556, + "traffic_rms": 0.07993554606098731 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.3501461160767576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5700354771813393e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00810072571424986, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17583886243343352 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10117068141698837, + "mean_feedback_forward_cosine": 0.9977079975989557, + "mean_feedback_forward_relative_error": 0.06663058481870159, + "min_feedback_forward_cosine": 0.9949249029159546 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036686295973838087, + "instruction_rms": 0.0036686295966257545, + "raw_apical_rms": 0.08000519481746854, + "teaching_rms": 0.0036686295973838087, + "traffic_rms": 0.0799184972858601 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.58910883843555e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5626243335983284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008036996723027936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.18267297471284866 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09699901938438416, + "mean_feedback_forward_cosine": 0.9978785880150334, + "mean_feedback_forward_relative_error": 0.06410341475519442, + "min_feedback_forward_cosine": 0.9953319430351257 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035932545201262144, + "instruction_rms": 0.003593254521193034, + "raw_apical_rms": 0.08042419208411777, + "teaching_rms": 0.0035932545201262144, + "traffic_rms": 0.08034153862534027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.108900514803281e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568811552101281e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007803228530432203, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18235203879356385 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09305055439472198, + "mean_feedback_forward_cosine": 0.9980383707631019, + "mean_feedback_forward_relative_error": 0.06164170729537164, + "min_feedback_forward_cosine": 0.9957046508789062 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036570393525167592, + "instruction_rms": 0.0036570393522322698, + "raw_apical_rms": 0.08090006371234526, + "teaching_rms": 0.0036570393525167592, + "traffic_rms": 0.08081434260689553 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.990508107814094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.566782802844155e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007849882222665925, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17558390679836272 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08932632952928543, + "mean_feedback_forward_cosine": 0.9981871362655393, + "mean_feedback_forward_relative_error": 0.05926036936861853, + "min_feedback_forward_cosine": 0.9960409998893738 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036348549668665927, + "instruction_rms": 0.003634854966830238, + "raw_apical_rms": 0.08086789785879021, + "teaching_rms": 0.0036348549668665927, + "traffic_rms": 0.08078358758754098 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.1231397912279135e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5693692218065148e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00787932761656052, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17688477809906006 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08556574583053589, + "mean_feedback_forward_cosine": 0.9983308930550853, + "mean_feedback_forward_relative_error": 0.05686707265915409, + "min_feedback_forward_cosine": 0.996366024017334 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036824688423622194, + "instruction_rms": 0.0036824688419950613, + "raw_apical_rms": 0.08102194083883581, + "teaching_rms": 0.0036824688423622194, + "traffic_rms": 0.08093564399966643 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.7526508123164604e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5685122866473856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008162393779176685, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.18277986051559447 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08226417750120163, + "mean_feedback_forward_cosine": 0.998453059504109, + "mean_feedback_forward_relative_error": 0.05474053779917379, + "min_feedback_forward_cosine": 0.996640145778656 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035474776656681615, + "instruction_rms": 0.003547477666022126, + "raw_apical_rms": 0.08132771503496944, + "teaching_rms": 0.0035474776656681615, + "traffic_rms": 0.08124752151904503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.334932685561398e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567679636746252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00774254198216927, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17239945264816284 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07894986122846603, + "mean_feedback_forward_cosine": 0.9985676811587426, + "mean_feedback_forward_relative_error": 0.05267049402238861, + "min_feedback_forward_cosine": 0.9969053268432617 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00356276355842355, + "instruction_rms": 0.0035627635580226906, + "raw_apical_rms": 0.0813899987839679, + "teaching_rms": 0.00356276355842355, + "traffic_rms": 0.08130950023272278 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7319087436694645e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5683558796275105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007647961973077621, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17563738648414612 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07576613873243332, + "mean_feedback_forward_cosine": 0.9986779305242723, + "mean_feedback_forward_relative_error": 0.05060064486197887, + "min_feedback_forward_cosine": 0.9971492290496826 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003669309407827526, + "instruction_rms": 0.003669309408464483, + "raw_apical_rms": 0.08158544149180676, + "teaching_rms": 0.003669309407827526, + "traffic_rms": 0.08150000676470577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.5702520051472675e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562578511492428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007938711460966998, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1760140047597885 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07266905903816223, + "mean_feedback_forward_cosine": 0.9987785816192627, + "mean_feedback_forward_relative_error": 0.04863823535701921, + "min_feedback_forward_cosine": 0.9973772764205933 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036194968051145686, + "instruction_rms": 0.003619496805162315, + "raw_apical_rms": 0.08191629950045785, + "teaching_rms": 0.0036194968051145686, + "traffic_rms": 0.08183385557538642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.452686669902505e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570160577014126e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00806506302170734, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17702248958110808 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06987287849187851, + "mean_feedback_forward_cosine": 0.998868546178264, + "mean_feedback_forward_relative_error": 0.04681353325084332, + "min_feedback_forward_cosine": 0.9975741505622864 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003575344953220973, + "instruction_rms": 0.0035753449538177234, + "raw_apical_rms": 0.0819683022658836, + "teaching_rms": 0.003575344953220973, + "traffic_rms": 0.08188777785956977 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.7679661229740304e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.566094433936314e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007849995753008054, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16953355858802796 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06717978417873383, + "mean_feedback_forward_cosine": 0.9989532828330994, + "mean_feedback_forward_relative_error": 0.045024896040558815, + "min_feedback_forward_cosine": 0.9977569580078125 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036141092943761297, + "instruction_rms": 0.003614109294875601, + "raw_apical_rms": 0.082077754216964, + "teaching_rms": 0.0036141092943761297, + "traffic_rms": 0.08199510688311354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.345369847806978e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562458871806964e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007238432633833196, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16995401079177858 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06453754007816315, + "mean_feedback_forward_cosine": 0.9990309669125464, + "mean_feedback_forward_relative_error": 0.0433202795383911, + "min_feedback_forward_cosine": 0.9979297518730164 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036439214262897262, + "instruction_rms": 0.003643921425967304, + "raw_apical_rms": 0.08217002811381474, + "teaching_rms": 0.0036439214262897262, + "traffic_rms": 0.08208661817354707 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0114140514569954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5690953303459456e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008072200522763718, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16821364754676818 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06191731244325638, + "mean_feedback_forward_cosine": 0.9991049516585565, + "mean_feedback_forward_relative_error": 0.04163601976488867, + "min_feedback_forward_cosine": 0.9980939626693726 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003617235617158337, + "instruction_rms": 0.0036172356182459903, + "raw_apical_rms": 0.08237902034936245, + "teaching_rms": 0.003617235617158337, + "traffic_rms": 0.08229712481692207 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.17764232704576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5625488105864348e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008154821467992599, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.17242857617855073 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059388451278209686, + "mean_feedback_forward_cosine": 0.9991734739272825, + "mean_feedback_forward_relative_error": 0.04001213655236267, + "min_feedback_forward_cosine": 0.998245894908905 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003620544539172883, + "instruction_rms": 0.0036205445387079436, + "raw_apical_rms": 0.08254011325402648, + "teaching_rms": 0.003620544539172883, + "traffic_rms": 0.08245780283950736 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.105209600562375e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.569397286008748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008291227381989149, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.1736736896419525 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.057012882083654404, + "mean_feedback_forward_cosine": 0.9992360165042262, + "mean_feedback_forward_relative_error": 0.03846835417132224, + "min_feedback_forward_cosine": 0.998383104801178 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035446815138317316, + "instruction_rms": 0.003544681514516423, + "raw_apical_rms": 0.0825348283111965, + "teaching_rms": 0.0035446815138317316, + "traffic_rms": 0.08245560063809228 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.697550272046101e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5656655588604398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00805729669538207, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1686829930496216 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.054785773158073425, + "mean_feedback_forward_cosine": 0.9992925120938209, + "mean_feedback_forward_relative_error": 0.03701847809697351, + "min_feedback_forward_cosine": 0.9985069632530212 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035615207358348114, + "instruction_rms": 0.00356152073636778, + "raw_apical_rms": 0.08277780108793044, + "teaching_rms": 0.0035615207358348114, + "traffic_rms": 0.08269831200569826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5986578222614298e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5606963094361545e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008063271781542992, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16682356884479524 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05265459045767784, + "mean_feedback_forward_cosine": 0.9993455429230967, + "mean_feedback_forward_relative_error": 0.035602630987282724, + "min_feedback_forward_cosine": 0.9986201524734497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00355541904800579, + "instruction_rms": 0.0035554190472763215, + "raw_apical_rms": 0.08282919608950909, + "teaching_rms": 0.00355541904800579, + "traffic_rms": 0.0827501483493525 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.679945237422544e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563713042214638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008130191503575137, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16671095200538635 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05064012482762337, + "mean_feedback_forward_cosine": 0.9993940841767096, + "mean_feedback_forward_relative_error": 0.03425592977193094, + "min_feedback_forward_cosine": 0.9987236261367798 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035852579973103065, + "instruction_rms": 0.003585257997799481, + "raw_apical_rms": 0.0830326928470821, + "teaching_rms": 0.0035852579973103065, + "traffic_rms": 0.08295268810956775 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.214773514059743e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5661120915691846e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008121393080230268, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16324914855003356 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048641324043273926, + "mean_feedback_forward_cosine": 0.9994407103907678, + "mean_feedback_forward_relative_error": 0.032912712393989484, + "min_feedback_forward_cosine": 0.998822033405304 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003544825413088148, + "instruction_rms": 0.003544825413466633, + "raw_apical_rms": 0.08327067909481489, + "teaching_rms": 0.003544825413088148, + "traffic_rms": 0.083192579049649 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.594163899833802e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5644479609335364e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00812012055725399, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1663533552932739 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04659680649638176, + "mean_feedback_forward_cosine": 0.9994832815662507, + "mean_feedback_forward_relative_error": 0.03163499982967492, + "min_feedback_forward_cosine": 0.9989189505577087 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003592078893184251, + "instruction_rms": 0.003592078893311404, + "raw_apical_rms": 0.0835031994484873, + "teaching_rms": 0.003592078893184251, + "traffic_rms": 0.08342315793935176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.235513901245213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5698272849513235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008072326017401142, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1693824523496628 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04479965567588806, + "mean_feedback_forward_cosine": 0.9995210882156126, + "mean_feedback_forward_relative_error": 0.03045479109090182, + "min_feedback_forward_cosine": 0.9990005493164062 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035323107009127327, + "instruction_rms": 0.003532310700250599, + "raw_apical_rms": 0.0837014063717534, + "teaching_rms": 0.0035323107009127327, + "traffic_rms": 0.08362399676257729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8977208188507575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5683451276094962e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007604283250805687, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16243568373680115 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04293741285800934, + "mean_feedback_forward_cosine": 0.9995575220354141, + "mean_feedback_forward_relative_error": 0.029275535185250542, + "min_feedback_forward_cosine": 0.9990818500518799 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003549721825434649, + "instruction_rms": 0.003549721825316782, + "raw_apical_rms": 0.08398704230406949, + "teaching_rms": 0.003549721825434649, + "traffic_rms": 0.08390917346887199 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.626586947940268e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5665325463905194e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00752577013343084, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16565432056427 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.041176553815603256, + "mean_feedback_forward_cosine": 0.999590939091098, + "mean_feedback_forward_relative_error": 0.028149035218502243, + "min_feedback_forward_cosine": 0.9991556406021118 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003576186377063354, + "instruction_rms": 0.0035761863772387054, + "raw_apical_rms": 0.08395016332387509, + "teaching_rms": 0.003576186377063354, + "traffic_rms": 0.0838712939257716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8516311090243107e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.569087439647886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008318848952305086, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16255624423503875 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.039520446211099625, + "mean_feedback_forward_cosine": 0.9996212951598629, + "mean_feedback_forward_relative_error": 0.0270841752269095, + "min_feedback_forward_cosine": 0.9992220401763916 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003561744843484287, + "instruction_rms": 0.003561744843839171, + "raw_apical_rms": 0.08399029340799966, + "teaching_rms": 0.003561744843484287, + "traffic_rms": 0.08391172999477906 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9455968331149052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5623253241720107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008006618363844678, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16491941719055175 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03796082362532616, + "mean_feedback_forward_cosine": 0.9996502822445285, + "mean_feedback_forward_relative_error": 0.026027433454029022, + "min_feedback_forward_cosine": 0.9992821216583252 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003589966664323821, + "instruction_rms": 0.003589966665122185, + "raw_apical_rms": 0.08431287707230326, + "teaching_rms": 0.003589966664323821, + "traffic_rms": 0.08423409906424048 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5056907659859462e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5609504358741697e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00785738037047338, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16296563831329347 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.036444567143917084, + "mean_feedback_forward_cosine": 0.9996770716482594, + "mean_feedback_forward_relative_error": 0.025009731011044596, + "min_feedback_forward_cosine": 0.9993383288383484 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00360747707542843, + "instruction_rms": 0.0036074770749768173, + "raw_apical_rms": 0.08467765755208641, + "teaching_rms": 0.00360747707542843, + "traffic_rms": 0.08459822310049236 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4535946607784354e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5648173371330087e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007617060488596373, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1663270664405823 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034938473254442215, + "mean_feedback_forward_cosine": 0.9997021286718307, + "mean_feedback_forward_relative_error": 0.02402060641156089, + "min_feedback_forward_cosine": 0.9993917942047119 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036397288463450774, + "instruction_rms": 0.0036397288465603184, + "raw_apical_rms": 0.08491326886517161, + "teaching_rms": 0.0036397288463450774, + "traffic_rms": 0.08483228167999984 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3745101884742326e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5604579782151287e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008048356497771771, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16660313208580016 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033616967499256134, + "mean_feedback_forward_cosine": 0.9997240112673852, + "mean_feedback_forward_relative_error": 0.023120274016213034, + "min_feedback_forward_cosine": 0.9994367361068726 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035727649303314837, + "instruction_rms": 0.003572764929611883, + "raw_apical_rms": 0.08470708676077847, + "teaching_rms": 0.0035727649303314837, + "traffic_rms": 0.08462902809964254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3300563256611895e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5630175216782487e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007916083632707675, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.15746098147869111 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03228490427136421, + "mean_feedback_forward_cosine": 0.9997451786072024, + "mean_feedback_forward_relative_error": 0.022217755431249257, + "min_feedback_forward_cosine": 0.9994805455207825 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003591549406742129, + "instruction_rms": 0.0035915494060704215, + "raw_apical_rms": 0.08518187075714867, + "teaching_rms": 0.003591549406742129, + "traffic_rms": 0.08510350021673546 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4713497026596087e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5628718047862487e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008340347177316647, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.16117573259353637 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032274410128593445, + "mean_feedback_forward_cosine": 0.9997453535756757, + "mean_feedback_forward_relative_error": 0.022208740619281606, + "min_feedback_forward_cosine": 0.9994809627532959 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025597685253738634, + "instruction_rms": 0.0025597685252116685, + "raw_apical_rms": 0.08547491858540754, + "teaching_rms": 0.0025597685253738634, + "traffic_rms": 0.08543433831941968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.13239105764751e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5494181237097096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007422654681911069, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09224812744140624 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03226881101727486, + "mean_feedback_forward_cosine": 0.999745447789469, + "mean_feedback_forward_relative_error": 0.022202638173175436, + "min_feedback_forward_cosine": 0.9994810819625854 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002305256666105552, + "instruction_rms": 0.0023052566671253245, + "raw_apical_rms": 0.08550807596709387, + "teaching_rms": 0.002305256666105552, + "traffic_rms": 0.08547501450326687 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.381442681416047e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5504861717517015e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007509226383192996, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.0670065698492527 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03226366266608238, + "mean_feedback_forward_cosine": 0.999745540080532, + "mean_feedback_forward_relative_error": 0.022197081797545957, + "min_feedback_forward_cosine": 0.999481201171875 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0022286802449406496, + "instruction_rms": 0.0022286802453555907, + "raw_apical_rms": 0.08545679273831264, + "teaching_rms": 0.0022286802449406496, + "traffic_rms": 0.08542551665802085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.29503713203394e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555665529858451e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007591604205528983, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05860980985045433 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03225874900817871, + "mean_feedback_forward_cosine": 0.9997456342943253, + "mean_feedback_forward_relative_error": 0.022191746340643977, + "min_feedback_forward_cosine": 0.9994813203811646 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002167464340917165, + "instruction_rms": 0.002167464341120168, + "raw_apical_rms": 0.08550673968990274, + "teaching_rms": 0.002167464340917165, + "traffic_rms": 0.08547679728379944 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8308521748954263e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5576558381197946e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007576745409090194, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.05120233317017555 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032253555953502655, + "mean_feedback_forward_cosine": 0.9997457285081187, + "mean_feedback_forward_relative_error": 0.02218653321746857, + "min_feedback_forward_cosine": 0.9994815587997437 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021619336690177904, + "instruction_rms": 0.00216193366919149, + "raw_apical_rms": 0.08544817313330487, + "teaching_rms": 0.0021619336690177904, + "traffic_rms": 0.08541816208300582 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.177058616188333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.554723711855487e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007505560624599145, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04782823716163635 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03224891051650047, + "mean_feedback_forward_cosine": 0.9997458246446425, + "mean_feedback_forward_relative_error": 0.022181579978355476, + "min_feedback_forward_cosine": 0.9994816780090332 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00208157284433858, + "instruction_rms": 0.0020815728441149596, + "raw_apical_rms": 0.08538747474352967, + "teaching_rms": 0.00208157284433858, + "traffic_rms": 0.08535949316124698 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6068104119758573e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5535638404695628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007590667119955549, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.043633259674310684 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03224467113614082, + "mean_feedback_forward_cosine": 0.999745899631131, + "mean_feedback_forward_relative_error": 0.022176929599335118, + "min_feedback_forward_cosine": 0.9994817972183228 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019755878934773264, + "instruction_rms": 0.0019755878929222856, + "raw_apical_rms": 0.0853358083673315, + "teaching_rms": 0.0019755878934773264, + "traffic_rms": 0.08531011942267654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.1214372480926613e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5580603677231063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007731737408741131, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.03839781272113323 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03224049508571625, + "mean_feedback_forward_cosine": 0.9997459707721588, + "mean_feedback_forward_relative_error": 0.02217231109558094, + "min_feedback_forward_cosine": 0.9994819164276123 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019690552945302426, + "instruction_rms": 0.0019690552947943503, + "raw_apical_rms": 0.08529715622856454, + "teaching_rms": 0.0019690552945302426, + "traffic_rms": 0.08527128827566889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.015750493773339e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558929068747043e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007791679900642506, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.036135771827697755 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032236065715551376, + "mean_feedback_forward_cosine": 0.9997460265313426, + "mean_feedback_forward_relative_error": 0.022167649931244312, + "min_feedback_forward_cosine": 0.9994820356369019 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020155333818530977, + "instruction_rms": 0.0020155333817156542, + "raw_apical_rms": 0.08530012725021925, + "teaching_rms": 0.0020155333818530977, + "traffic_rms": 0.08527326971106068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.6608518882985663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5544662176204826e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007744373681210149, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03552405541539192 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03223174065351486, + "mean_feedback_forward_cosine": 0.9997461380497101, + "mean_feedback_forward_relative_error": 0.022162993814075185, + "min_feedback_forward_cosine": 0.9994823336601257 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019946874885367533, + "instruction_rms": 0.001994687488528825, + "raw_apical_rms": 0.085235179910735, + "teaching_rms": 0.0019946874885367533, + "traffic_rms": 0.08520812312264033 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.1490266394484934e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5534802898873948e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007736756121215812, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.032745867322683336 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032227423042058945, + "mean_feedback_forward_cosine": 0.9997462149589292, + "mean_feedback_forward_relative_error": 0.02215854586252282, + "min_feedback_forward_cosine": 0.9994823932647705 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019275506545169381, + "instruction_rms": 0.001927550654109393, + "raw_apical_rms": 0.08522787394399604, + "teaching_rms": 0.0019275506545169381, + "traffic_rms": 0.08520268849049957 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.212632648796898e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552632422780837e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007719882621417421, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.030830372276902197 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03222312778234482, + "mean_feedback_forward_cosine": 0.9997463091727226, + "mean_feedback_forward_relative_error": 0.02215408224372133, + "min_feedback_forward_cosine": 0.9994825124740601 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00196718213783514, + "instruction_rms": 0.0019671821376934396, + "raw_apical_rms": 0.08521148228723689, + "teaching_rms": 0.00196718213783514, + "traffic_rms": 0.08518520176123512 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8163565701788684e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5575391596702754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007679841139779355, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.029900190619826315 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032219015061855316, + "mean_feedback_forward_cosine": 0.9997463937728636, + "mean_feedback_forward_relative_error": 0.022149722482408245, + "min_feedback_forward_cosine": 0.9994826316833496 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018982169025729806, + "instruction_rms": 0.0018982169027630233, + "raw_apical_rms": 0.08521080114565163, + "teaching_rms": 0.0018982169025729806, + "traffic_rms": 0.08518592287703344 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7031020977119624e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5574630539534517e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007789678385508049, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.02714966125845909 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03221457079052925, + "mean_feedback_forward_cosine": 0.9997464495320474, + "mean_feedback_forward_relative_error": 0.022145199754665935, + "min_feedback_forward_cosine": 0.9994828104972839 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00194397035899198, + "instruction_rms": 0.0019439703589415216, + "raw_apical_rms": 0.0851888517811902, + "teaching_rms": 0.00194397035899198, + "traffic_rms": 0.08516212544480485 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.873636262279433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5565259955891475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007707835169791278, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.028146579477787016 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03221031650900841, + "mean_feedback_forward_cosine": 0.9997465572049541, + "mean_feedback_forward_relative_error": 0.02214083282817756, + "min_feedback_forward_cosine": 0.999483048915863 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018825244186822523, + "instruction_rms": 0.001882524418313962, + "raw_apical_rms": 0.08517330050297331, + "teaching_rms": 0.0018825244186822523, + "traffic_rms": 0.08514870388517795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.462081027942986e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556036031876749e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007815636675572953, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.02523015166759491 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032206013798713684, + "mean_feedback_forward_cosine": 0.9997466110414074, + "mean_feedback_forward_relative_error": 0.02213649634992884, + "min_feedback_forward_cosine": 0.9994831085205078 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018542901978201884, + "instruction_rms": 0.0018542901979495385, + "raw_apical_rms": 0.08514926068159522, + "teaching_rms": 0.0018542901978201884, + "traffic_rms": 0.0851253207030986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.7652473139360154e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550396968613429e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0077644035624857485, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024127442442178727 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03220193088054657, + "mean_feedback_forward_cosine": 0.9997466975642789, + "mean_feedback_forward_relative_error": 0.022132312172963735, + "min_feedback_forward_cosine": 0.9994832873344421 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017883914950069214, + "instruction_rms": 0.0017883914951428513, + "raw_apical_rms": 0.0851536953671786, + "teaching_rms": 0.0017883914950069214, + "traffic_rms": 0.08513118394129957 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.743103166835681e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5570751830994022e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007698554395149485, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.02251512383878231 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03219754993915558, + "mean_feedback_forward_cosine": 0.9997467802416894, + "mean_feedback_forward_relative_error": 0.02212798718603388, + "min_feedback_forward_cosine": 0.9994834065437317 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001872706368293738, + "instruction_rms": 0.0018727063690515402, + "raw_apical_rms": 0.085143720624432, + "teaching_rms": 0.001872706368293738, + "traffic_rms": 0.08511863599068728 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8432380005938285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.557123742589292e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007778052383629886, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.023637037647366525 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03219319134950638, + "mean_feedback_forward_cosine": 0.9997468667645608, + "mean_feedback_forward_relative_error": 0.02212357660755515, + "min_feedback_forward_cosine": 0.9994834661483765 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018306515800131302, + "instruction_rms": 0.0018306515801432503, + "raw_apical_rms": 0.08514329400720481, + "teaching_rms": 0.0018306515800131302, + "traffic_rms": 0.085119184102793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.7137616598338354e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552462742484363e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007717499999142802, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.022211281167566776 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03218923509120941, + "mean_feedback_forward_cosine": 0.9997469436737799, + "mean_feedback_forward_relative_error": 0.022119513096948786, + "min_feedback_forward_cosine": 0.9994837045669556 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017249871914953483, + "instruction_rms": 0.0017249871909884261, + "raw_apical_rms": 0.08509293233976999, + "teaching_rms": 0.0017249871914953483, + "traffic_rms": 0.08507107952327729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5879602461031936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5573660121037202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007818782947563609, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.020295898990035057 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03218502178788185, + "mean_feedback_forward_cosine": 0.9997470225057294, + "mean_feedback_forward_relative_error": 0.022115399490200704, + "min_feedback_forward_cosine": 0.9994838237762451 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017326894230005753, + "instruction_rms": 0.001732689423198971, + "raw_apical_rms": 0.08506545722100739, + "teaching_rms": 0.0017326894230005753, + "traffic_rms": 0.08504346085471128 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0309909249126576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562044051878452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007684725353176205, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.02001648464664817 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0321810320019722, + "mean_feedback_forward_cosine": 0.9997471051831399, + "mean_feedback_forward_relative_error": 0.02211112853500151, + "min_feedback_forward_cosine": 0.9994838237762451 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001750478105159913, + "instruction_rms": 0.001750478104743195, + "raw_apical_rms": 0.0850260850495229, + "teaching_rms": 0.001750478105159913, + "traffic_rms": 0.08500283254868161 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7447979888961527e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555106975096423e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007693639945793241, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01908622027337551 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032176803797483444, + "mean_feedback_forward_cosine": 0.999747182092359, + "mean_feedback_forward_relative_error": 0.022106929036277916, + "min_feedback_forward_cosine": 0.9994839429855347 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017197131457769262, + "instruction_rms": 0.0017197131461424737, + "raw_apical_rms": 0.08506510089685689, + "teaching_rms": 0.0017197131457769262, + "traffic_rms": 0.08504328450881858 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.1180100279470935e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556117684510507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007912134930753045, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018922930538356305 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0321723110973835, + "mean_feedback_forward_cosine": 0.9997472647697695, + "mean_feedback_forward_relative_error": 0.022102614789600333, + "min_feedback_forward_cosine": 0.9994841814041138 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001724766291442829, + "instruction_rms": 0.0017247662909012323, + "raw_apical_rms": 0.08505729341463186, + "teaching_rms": 0.001724766291442829, + "traffic_rms": 0.0850345928067812 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.171100004806249e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5614709588820308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00776633463105838, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.01882740825429559 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03216816857457161, + "mean_feedback_forward_cosine": 0.9997473609062933, + "mean_feedback_forward_relative_error": 0.022098499094887127, + "min_feedback_forward_cosine": 0.9994843006134033 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016975205727734217, + "instruction_rms": 0.0016975205725969845, + "raw_apical_rms": 0.08499906278563381, + "teaching_rms": 0.0016975205727734217, + "traffic_rms": 0.08497725852671685 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.053553854215352e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5609368579527885e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007723970167144092, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01734948897063732 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032163869589567184, + "mean_feedback_forward_cosine": 0.999747432047321, + "mean_feedback_forward_relative_error": 0.022094277875317682, + "min_feedback_forward_cosine": 0.9994844794273376 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00170715901790463, + "instruction_rms": 0.0017071590181106844, + "raw_apical_rms": 0.0850190448271698, + "teaching_rms": 0.00170715901790463, + "traffic_rms": 0.08499731895421725 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.214280397401126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5570650276014787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007805124050160235, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01757803070306778 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03215986117720604, + "mean_feedback_forward_cosine": 0.9997475147247314, + "mean_feedback_forward_relative_error": 0.0220902826516859, + "min_feedback_forward_cosine": 0.9994845390319824 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015933217839835534, + "instruction_rms": 0.0015933217835486816, + "raw_apical_rms": 0.08497799004044053, + "teaching_rms": 0.0015933217839835534, + "traffic_rms": 0.08495892047212655 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.829875889176492e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5560828380794952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008006805089130596, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.016096085978746413 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03215549886226654, + "mean_feedback_forward_cosine": 0.9997475858657591, + "mean_feedback_forward_relative_error": 0.022086039741313266, + "min_feedback_forward_cosine": 0.9994845986366272 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016593744626707601, + "instruction_rms": 0.0016593744631088348, + "raw_apical_rms": 0.08495291697640645, + "teaching_rms": 0.0016593744626707601, + "traffic_rms": 0.08493118008371006 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5495866751157337e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.560079368416829e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007859884443728348, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01627414563268423 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03215118497610092, + "mean_feedback_forward_cosine": 0.999747674311361, + "mean_feedback_forward_relative_error": 0.022081790882493218, + "min_feedback_forward_cosine": 0.9994847774505615 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016172870188036102, + "instruction_rms": 0.0016172870183968839, + "raw_apical_rms": 0.0849784783314686, + "teaching_rms": 0.0016172870188036102, + "traffic_rms": 0.08495804235749042 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7968557247959325e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5541551129593037e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007806808931482821, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.015499272239804268 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03214675933122635, + "mean_feedback_forward_cosine": 0.9997477454523886, + "mean_feedback_forward_relative_error": 0.02207767592382527, + "min_feedback_forward_cosine": 0.9994849562644958 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001587400652656276, + "instruction_rms": 0.0015874006521807446, + "raw_apical_rms": 0.08494472464965593, + "teaching_rms": 0.001587400652656276, + "traffic_rms": 0.08492448017847222 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6374522121841437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5548840015389928e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007799769111020006, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.015062347439229488 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032142650336027145, + "mean_feedback_forward_cosine": 0.9997478492798344, + "mean_feedback_forward_relative_error": 0.022073371290800073, + "min_feedback_forward_cosine": 0.9994851350784302 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016659894306172687, + "instruction_rms": 0.0016659894307567659, + "raw_apical_rms": 0.08493607010966901, + "teaching_rms": 0.0016659894306172687, + "traffic_rms": 0.08491422903213676 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5005007842082705e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.553253938898014e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00784193991975821, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.016026172676235437 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032138071954250336, + "mean_feedback_forward_cosine": 0.9997479281117839, + "mean_feedback_forward_relative_error": 0.022068944414176287, + "min_feedback_forward_cosine": 0.9994852542877197 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016422839961681608, + "instruction_rms": 0.0016422839962792384, + "raw_apical_rms": 0.08490657511741903, + "teaching_rms": 0.0016422839961681608, + "traffic_rms": 0.08488532509625313 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.400448195771787e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555470969678772e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007999214170619868, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.015466258705556393 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03213370218873024, + "mean_feedback_forward_cosine": 0.9997480146346553, + "mean_feedback_forward_relative_error": 0.0220646015368402, + "min_feedback_forward_cosine": 0.9994853734970093 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001621059182814861, + "instruction_rms": 0.0016210591822908918, + "raw_apical_rms": 0.08487005716680976, + "teaching_rms": 0.001621059182814861, + "traffic_rms": 0.08484914019158334 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5937458403291203e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5620507895283276e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007905574689853619, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.01449645267546177 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03212935850024223, + "mean_feedback_forward_cosine": 0.9997480992347964, + "mean_feedback_forward_relative_error": 0.022060277736595562, + "min_feedback_forward_cosine": 0.9994856119155884 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015873318508634684, + "instruction_rms": 0.0015873318504414258, + "raw_apical_rms": 0.08485106889697824, + "teaching_rms": 0.0015873318508634684, + "traffic_rms": 0.0848306373081642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.842952007142685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5554440319659977e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007860545566643501, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.014009355556964875 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032125324010849, + "mean_feedback_forward_cosine": 0.999748170375824, + "mean_feedback_forward_relative_error": 0.022056281761897188, + "min_feedback_forward_cosine": 0.9994857311248779 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015373935054857053, + "instruction_rms": 0.0015373935059749735, + "raw_apical_rms": 0.0848535883891063, + "teaching_rms": 0.0015373935054857053, + "traffic_rms": 0.08483520991396627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.0560477764774987e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5556613977708963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00806323094107874, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013058475132286548 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03212100639939308, + "mean_feedback_forward_cosine": 0.9997482761260001, + "mean_feedback_forward_relative_error": 0.022052032121967886, + "min_feedback_forward_cosine": 0.9994857907295227 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015374782100378457, + "instruction_rms": 0.0015374782099927705, + "raw_apical_rms": 0.08480027915978967, + "teaching_rms": 0.0015374782100378457, + "traffic_rms": 0.08478078771374715 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.649736963178384e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5547951667510203e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008010708154023244, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.013404304009526968 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03211704641580582, + "mean_feedback_forward_cosine": 0.9997483357306449, + "mean_feedback_forward_relative_error": 0.02204796293329808, + "min_feedback_forward_cosine": 0.9994858503341675 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015169432487520222, + "instruction_rms": 0.0015169432489392186, + "raw_apical_rms": 0.08481512105806807, + "teaching_rms": 0.0015169432487520222, + "traffic_rms": 0.08479604233934757 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7692978576965288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5548446199935154e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007842710172121984, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.012942820435762406 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032112620770931244, + "mean_feedback_forward_cosine": 0.9997484472490126, + "mean_feedback_forward_relative_error": 0.022043740241638115, + "min_feedback_forward_cosine": 0.9994860291481018 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014693324340644335, + "instruction_rms": 0.0014693324339544657, + "raw_apical_rms": 0.08477673160035112, + "teaching_rms": 0.0014693324340644335, + "traffic_rms": 0.08475895242745539 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3616612665477987e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550393502031624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007822802409237734, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012175001892745495 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032108407467603683, + "mean_feedback_forward_cosine": 0.9997485222355011, + "mean_feedback_forward_relative_error": 0.02203956582853871, + "min_feedback_forward_cosine": 0.9994860887527466 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001475875172260604, + "instruction_rms": 0.0014758751723666435, + "raw_apical_rms": 0.08481288710500928, + "teaching_rms": 0.001475875172260604, + "traffic_rms": 0.08479507375553047 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.3507863150816775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.554567223700286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007895884904967414, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012583747244477273 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03210366889834404, + "mean_feedback_forward_cosine": 0.9997486106811031, + "mean_feedback_forward_relative_error": 0.02203505710067768, + "min_feedback_forward_cosine": 0.9994863271713257 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015424828635693577, + "instruction_rms": 0.0015424828637368346, + "raw_apical_rms": 0.08477559262783195, + "teaching_rms": 0.0015424828635693577, + "traffic_rms": 0.08475525228086862 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.998925907604644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.557466878811635e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007794211460095631, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012852763197124004 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03209934011101723, + "mean_feedback_forward_cosine": 0.9997486895130526, + "mean_feedback_forward_relative_error": 0.022030829587170193, + "min_feedback_forward_cosine": 0.9994864463806152 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014845357426710888, + "instruction_rms": 0.001484535742014835, + "raw_apical_rms": 0.08473327291199707, + "teaching_rms": 0.0014845357426710888, + "traffic_rms": 0.08471455306199036 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.648898783126242e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5535140471736675e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00784882976260942, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.011782860443741083 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032095324248075485, + "mean_feedback_forward_cosine": 0.9997487683450023, + "mean_feedback_forward_relative_error": 0.022026723461045374, + "min_feedback_forward_cosine": 0.99948650598526 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001427361962774986, + "instruction_rms": 0.0014273619625849627, + "raw_apical_rms": 0.08474727899816284, + "teaching_rms": 0.001427361962774986, + "traffic_rms": 0.08473037693091633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4972516499133235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5536132661164392e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007961414213969986, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011186469346061349 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032091110944747925, + "mean_feedback_forward_cosine": 0.9997488548678737, + "mean_feedback_forward_relative_error": 0.022022511870149645, + "min_feedback_forward_cosine": 0.9994868040084839 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014574241359160468, + "instruction_rms": 0.0014574241362217426, + "raw_apical_rms": 0.08474393105256815, + "teaching_rms": 0.0014574241359160468, + "traffic_rms": 0.08472569383105438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6983251188994473e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.554834923581291e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007946274459855015, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011221854297220706 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032086823135614395, + "mean_feedback_forward_cosine": 0.9997489510043975, + "mean_feedback_forward_relative_error": 0.022018264333206798, + "min_feedback_forward_cosine": 0.9994869232177734 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001448803525872259, + "instruction_rms": 0.0014488035262046378, + "raw_apical_rms": 0.08472024989030588, + "teaching_rms": 0.001448803525872259, + "traffic_rms": 0.08470302759221932 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9520295541374253e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5558101714371867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0081041679822775, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.01130846711948514 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032081812620162964, + "mean_feedback_forward_cosine": 0.999749033681808, + "mean_feedback_forward_relative_error": 0.022013545652190523, + "min_feedback_forward_cosine": 0.999487042427063 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001534790630143416, + "instruction_rms": 0.0015347906298445008, + "raw_apical_rms": 0.08477277706807212, + "teaching_rms": 0.001534790630143416, + "traffic_rms": 0.08475246828589694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.386415332011893e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5519213591166178e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008177826339476349, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.01187979559995234 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032077133655548096, + "mean_feedback_forward_cosine": 0.9997491394319842, + "mean_feedback_forward_relative_error": 0.02200902559824528, + "min_feedback_forward_cosine": 0.9994872808456421 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014727569943933216, + "instruction_rms": 0.0014727569941921865, + "raw_apical_rms": 0.08472413737652122, + "teaching_rms": 0.0014727569943933216, + "traffic_rms": 0.08470588662075917 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.4959424749226855e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5499015325663085e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008075416976138882, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.01102683356255293 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032072003930807114, + "mean_feedback_forward_cosine": 0.9997492374912385, + "mean_feedback_forward_relative_error": 0.022004058434357567, + "min_feedback_forward_cosine": 0.9994872808456421 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015278313777748094, + "instruction_rms": 0.0015278313778563416, + "raw_apical_rms": 0.08469980169741469, + "teaching_rms": 0.0015278313777748094, + "traffic_rms": 0.0846800555268072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.034852369410146e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556818837641004e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007988470625822102, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.012224403989054262 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03206689655780792, + "mean_feedback_forward_cosine": 0.9997493355504928, + "mean_feedback_forward_relative_error": 0.021999290813842127, + "min_feedback_forward_cosine": 0.9994876384735107 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015078581384401503, + "instruction_rms": 0.0015078581381856684, + "raw_apical_rms": 0.08465784205089255, + "teaching_rms": 0.0015078581384401503, + "traffic_rms": 0.08463773276682061 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4938001885743688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5604575214776513e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00822341808139963, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.0114341216224432 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03206200897693634, + "mean_feedback_forward_cosine": 0.9997494239960948, + "mean_feedback_forward_relative_error": 0.021994415820846634, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001445535362554618, + "instruction_rms": 0.0014455353626978988, + "raw_apical_rms": 0.08463006693227192, + "teaching_rms": 0.001445535362554618, + "traffic_rms": 0.08461104184464936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9224544562111987e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5588713403216154e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008134733931461751, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.010759009138792753 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032057419419288635, + "mean_feedback_forward_cosine": 0.9997495239780795, + "mean_feedback_forward_relative_error": 0.021990107777979103, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013580066174756673, + "instruction_rms": 0.0013580066174834781, + "raw_apical_rms": 0.08459505879010035, + "teaching_rms": 0.0013580066174756673, + "traffic_rms": 0.08457888214262231 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7953472425679138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5602074224459863e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008016322437811791, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009937263458222151 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205735981464386, + "mean_feedback_forward_cosine": 0.9997495028280443, + "mean_feedback_forward_relative_error": 0.021990005452666554, + "min_feedback_forward_cosine": 0.9994876980781555 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012989680606928906, + "instruction_rms": 0.001298968060843548, + "raw_apical_rms": 0.08462796473056299, + "teaching_rms": 0.0012989680606928906, + "traffic_rms": 0.08461294314685697 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.1304224989980867e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.553054856435117e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008106089138899994, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008963629547953605 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205730766057968, + "mean_feedback_forward_cosine": 0.999749529746271, + "mean_feedback_forward_relative_error": 0.021989917818216548, + "min_feedback_forward_cosine": 0.9994876384735107 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001293735707241519, + "instruction_rms": 0.0012937357074909064, + "raw_apical_rms": 0.08462942698258834, + "teaching_rms": 0.001293735707241519, + "traffic_rms": 0.08461440562495595 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.821867006730372e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5519608343091926e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008027482755381848, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008981580437123776 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032057277858257294, + "mean_feedback_forward_cosine": 0.9997495374371929, + "mean_feedback_forward_relative_error": 0.02198983958712028, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011501281662215492, + "instruction_rms": 0.001150128166318779, + "raw_apical_rms": 0.08461995237652827, + "teaching_rms": 0.0011501281662215492, + "traffic_rms": 0.08460733298068783 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.404548204992807e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558381581211012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008084671228267505, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008220292322188617 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032057248055934906, + "mean_feedback_forward_cosine": 0.9997495316690014, + "mean_feedback_forward_relative_error": 0.021989753665102106, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001184094101465768, + "instruction_rms": 0.0011840941014483431, + "raw_apical_rms": 0.08462711483998296, + "teaching_rms": 0.001184094101465768, + "traffic_rms": 0.0846143558336007 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6696923782765225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55720848437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008082900324625079, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.008421510046869517 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205721825361252, + "mean_feedback_forward_cosine": 0.9997495316690014, + "mean_feedback_forward_relative_error": 0.021989670116454363, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011230874228691688, + "instruction_rms": 0.0011230874226948217, + "raw_apical_rms": 0.08462608529514877, + "teaching_rms": 0.0011230874228691688, + "traffic_rms": 0.08461442307945517 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7876859931552644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5539234208608452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008063827890405058, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.007678777284361422 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205718472599983, + "mean_feedback_forward_cosine": 0.9997495085962357, + "mean_feedback_forward_relative_error": 0.021989584239500183, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011594367387658278, + "instruction_rms": 0.0011594367388436816, + "raw_apical_rms": 0.0846299918445453, + "teaching_rms": 0.0011594367387658278, + "traffic_rms": 0.08461697493736504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.24372599891376e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555038123594824e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008076651915115954, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.00785817270964384 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205714747309685, + "mean_feedback_forward_cosine": 0.99974952590081, + "mean_feedback_forward_relative_error": 0.021989505347465316, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011657245365337644, + "instruction_rms": 0.0011657245369507377, + "raw_apical_rms": 0.08463832061075031, + "teaching_rms": 0.0011657245365337644, + "traffic_rms": 0.0846253047091063 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.5746492937628318e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5575715595288017e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008105453680875676, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.008191546478867532 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205711394548416, + "mean_feedback_forward_cosine": 0.9997495393599233, + "mean_feedback_forward_relative_error": 0.02198942340610008, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011683804452726383, + "instruction_rms": 0.0011683804452906222, + "raw_apical_rms": 0.08463754597318865, + "teaching_rms": 0.0011683804452726383, + "traffic_rms": 0.0846246500734196 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6712719147781172e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5544190917292254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008005810984520589, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.00793969894580543 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032057080417871475, + "mean_feedback_forward_cosine": 0.999749529746271, + "mean_feedback_forward_relative_error": 0.021989342771590717, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001101375217821854, + "instruction_rms": 0.001101375217456913, + "raw_apical_rms": 0.08463425646961152, + "teaching_rms": 0.001101375217821854, + "traffic_rms": 0.08462316520222135 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.2625536050545634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.553113066484282e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008031031205818068, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007505376897156238 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205704316496849, + "mean_feedback_forward_cosine": 0.9997495451281148, + "mean_feedback_forward_relative_error": 0.021989260905332143, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001114028869247911, + "instruction_rms": 0.0011140288690131368, + "raw_apical_rms": 0.08463153749097074, + "teaching_rms": 0.001114028869247911, + "traffic_rms": 0.08461929183551309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.301671114286364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5513995964226713e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008035934509541286, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007374777395576238 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0320570170879364, + "mean_feedback_forward_cosine": 0.9997495374371929, + "mean_feedback_forward_relative_error": 0.021989184566923687, + "min_feedback_forward_cosine": 0.9994879961013794 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001078348365094008, + "instruction_rms": 0.001078348364999181, + "raw_apical_rms": 0.08462990362008894, + "teaching_rms": 0.001078348365094008, + "traffic_rms": 0.08461840161094894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6519327178675667e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5535282544872595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008031078324797637, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0071386242005229 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056983560323715, + "mean_feedback_forward_cosine": 0.9997495335917319, + "mean_feedback_forward_relative_error": 0.021989107222086, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001131726213357515, + "instruction_rms": 0.0011317262130276604, + "raw_apical_rms": 0.08462298193497658, + "teaching_rms": 0.001131726213357515, + "traffic_rms": 0.08461029880638411 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.904303642026207e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552688714270153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007990689282038173, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007564505198374391 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205695003271103, + "mean_feedback_forward_cosine": 0.9997495335917319, + "mean_feedback_forward_relative_error": 0.021989029862226978, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011427993318255945, + "instruction_rms": 0.0011427993320450444, + "raw_apical_rms": 0.08462768994713521, + "teaching_rms": 0.0011427993318255945, + "traffic_rms": 0.08461507765407264 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5662766134360027e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55643257860433e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008031436327362714, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0077449878285825256 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205691650509834, + "mean_feedback_forward_cosine": 0.9997495220553491, + "mean_feedback_forward_relative_error": 0.021988952156877326, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011349494656758055, + "instruction_rms": 0.0011349494659700836, + "raw_apical_rms": 0.08462813089448606, + "teaching_rms": 0.0011349494656758055, + "traffic_rms": 0.08461525609875001 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.057690634022947e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5599110211517814e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00796175083263991, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007448814112208784 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056886702775955, + "mean_feedback_forward_cosine": 0.9997495374371929, + "mean_feedback_forward_relative_error": 0.0219888728142025, + "min_feedback_forward_cosine": 0.9994876980781555 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010273276931902122, + "instruction_rms": 0.0010273276931739626, + "raw_apical_rms": 0.08462029847856628, + "teaching_rms": 0.0010273276931902122, + "traffic_rms": 0.08460980368635514 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7273271641092833e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5539829465520633e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007998241362975582, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.007065627899318933 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056860625743866, + "mean_feedback_forward_cosine": 0.9997495432053843, + "mean_feedback_forward_relative_error": 0.021988793546634335, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010886068279447245, + "instruction_rms": 0.001088606827894687, + "raw_apical_rms": 0.08462434490255452, + "teaching_rms": 0.0010886068279447245, + "traffic_rms": 0.08461260871484577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.588424408145329e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5559317753020518e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008060178923424253, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.00736294310901314 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205682709813118, + "mean_feedback_forward_cosine": 0.9997495508963062, + "mean_feedback_forward_relative_error": 0.021988717959292473, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010434216503069343, + "instruction_rms": 0.0010434216502617963, + "raw_apical_rms": 0.0846309903455389, + "teaching_rms": 0.0010434216503069343, + "traffic_rms": 0.08461957293538784 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.976225133011212e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5561108411374453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007937793166609995, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007084802482575178 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056793570518494, + "mean_feedback_forward_cosine": 0.9997495432053843, + "mean_feedback_forward_relative_error": 0.02198864073462544, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001087117930368285, + "instruction_rms": 0.0010871179301175132, + "raw_apical_rms": 0.0846282357229301, + "teaching_rms": 0.001087117930368285, + "traffic_rms": 0.08461685172046067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4137907672638903e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5560546328165952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007978622796526343, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.007348423615917563 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056767493486404, + "mean_feedback_forward_cosine": 0.9997495547417672, + "mean_feedback_forward_relative_error": 0.021988567836101976, + "min_feedback_forward_cosine": 0.9994876980781555 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010287706784016426, + "instruction_rms": 0.001028770678573291, + "raw_apical_rms": 0.08463001039004983, + "teaching_rms": 0.0010287706784016426, + "traffic_rms": 0.08461976576247446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.0807619566530775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555825872490855e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007986188546872202, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006677203074581921 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056741416454315, + "mean_feedback_forward_cosine": 0.9997495508963062, + "mean_feedback_forward_relative_error": 0.021988490956925576, + "min_feedback_forward_cosine": 0.9994876980781555 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001011536117134013, + "instruction_rms": 0.0010115361168249457, + "raw_apical_rms": 0.08463372515507794, + "teaching_rms": 0.001011536117134013, + "traffic_rms": 0.08462400595222735 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.4528063136131095e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5574455085669423e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007958210091112737, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0066659953045845035 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056715339422226, + "mean_feedback_forward_cosine": 0.9997495739690719, + "mean_feedback_forward_relative_error": 0.02198841607558631, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010305776914244968, + "instruction_rms": 0.0010305776920129862, + "raw_apical_rms": 0.0846315625766151, + "teaching_rms": 0.0010305776914244968, + "traffic_rms": 0.08462115907843928 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6995307322863482e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55666453851647e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007948714045190972, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.00684373818129301 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056692987680435, + "mean_feedback_forward_cosine": 0.9997495316690014, + "mean_feedback_forward_relative_error": 0.021988339947476503, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010185823995356566, + "instruction_rms": 0.0010185823997706453, + "raw_apical_rms": 0.08463682666297681, + "teaching_rms": 0.0010185823995356566, + "traffic_rms": 0.08462701372469258 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.10159400385146e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5578661381177768e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00799824641272426, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006583357010632753 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056670635938644, + "mean_feedback_forward_cosine": 0.9997495412826538, + "mean_feedback_forward_relative_error": 0.021988270068240744, + "min_feedback_forward_cosine": 0.9994879364967346 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010143742825028702, + "instruction_rms": 0.001014374282225172, + "raw_apical_rms": 0.08463549558025417, + "teaching_rms": 0.0010143742825028702, + "traffic_rms": 0.08462586229141997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5325071817405842e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5617552550658932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008056666391193368, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006385309305116534 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205663338303566, + "mean_feedback_forward_cosine": 0.99974956627815, + "mean_feedback_forward_relative_error": 0.02198819270838172, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001061901406817756, + "instruction_rms": 0.0010619014068187391, + "raw_apical_rms": 0.08463902423496301, + "teaching_rms": 0.001061901406817756, + "traffic_rms": 0.08462801164852572 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5313958929018554e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5591065058488818e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007962355973333154, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.007210728006362915 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205661103129387, + "mean_feedback_forward_cosine": 0.9997495566644976, + "mean_feedback_forward_relative_error": 0.0219881187433437, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009435954855030087, + "instruction_rms": 0.0009435954852273114, + "raw_apical_rms": 0.08463053724454965, + "teaching_rms": 0.0009435954855030087, + "traffic_rms": 0.08462131963923884 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.630564921446421e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5551755703600296e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00796815046325488, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006028526222705841 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205658495426178, + "mean_feedback_forward_cosine": 0.9997495374371929, + "mean_feedback_forward_relative_error": 0.021988044748263012, + "min_feedback_forward_cosine": 0.9994879364967346 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010496879982958467, + "instruction_rms": 0.0010496879984062887, + "raw_apical_rms": 0.08463434562890049, + "teaching_rms": 0.0010496879982958467, + "traffic_rms": 0.0846235661975684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 3.0546686957677593e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5561648529717188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008099627588135169, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006795201230123639 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205656260251999, + "mean_feedback_forward_cosine": 0.9997495778145329, + "mean_feedback_forward_relative_error": 0.02198797264587014, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000954790303047351, + "instruction_rms": 0.0009547903024134375, + "raw_apical_rms": 0.08463088628265412, + "teaching_rms": 0.000954790303047351, + "traffic_rms": 0.08462221980843948 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.6920155821462686e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5570191817071018e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008016009289826279, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006276105828657746 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0320565290749073, + "mean_feedback_forward_cosine": 0.9997495566644976, + "mean_feedback_forward_relative_error": 0.02198789587184306, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009742248533581468, + "instruction_rms": 0.0009742248532229731, + "raw_apical_rms": 0.08463012244590522, + "teaching_rms": 0.0009742248533581468, + "traffic_rms": 0.0846208089619147 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7126036528102304e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5541531310149866e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007964348409284004, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.006347799450755119 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205651044845581, + "mean_feedback_forward_cosine": 0.9997495643554195, + "mean_feedback_forward_relative_error": 0.02198782029952253, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009686657426061795, + "instruction_rms": 0.0009686657424796098, + "raw_apical_rms": 0.08463348324278785, + "teaching_rms": 0.0009686657426061795, + "traffic_rms": 0.08462395844964873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5305604722802855e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558070982569098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008004684243068633, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0063541915854811665 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056476920843124, + "mean_feedback_forward_cosine": 0.999749570123611, + "mean_feedback_forward_relative_error": 0.02198774262421554, + "min_feedback_forward_cosine": 0.9994879364967346 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010610018039601341, + "instruction_rms": 0.0010610018039791782, + "raw_apical_rms": 0.08463786018054677, + "teaching_rms": 0.0010610018039601341, + "traffic_rms": 0.08462636198447633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.991505354543733e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5552112802124914e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007951876528440281, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.00676737730845809 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205644711852074, + "mean_feedback_forward_cosine": 0.999749570123611, + "mean_feedback_forward_relative_error": 0.021987671678465222, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009837532117186695, + "instruction_rms": 0.0009837532114438997, + "raw_apical_rms": 0.08463568361682981, + "teaching_rms": 0.0009837532117186695, + "traffic_rms": 0.08462597539949036 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3788350844711204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5557195198864044e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008089788455360023, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006207149089500308 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205642104148865, + "mean_feedback_forward_cosine": 0.9997495585872281, + "mean_feedback_forward_relative_error": 0.02198760001169097, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009525988091816854, + "instruction_rms": 0.0009525988097080265, + "raw_apical_rms": 0.08463267729022368, + "teaching_rms": 0.0009525988091816854, + "traffic_rms": 0.08462361191370187 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.7561715515721517e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5581113218914804e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007955770265707164, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.005954133263677359 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205639496445656, + "mean_feedback_forward_cosine": 0.9997495470508453, + "mean_feedback_forward_relative_error": 0.021987526256951592, + "min_feedback_forward_cosine": 0.9994879364967346 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010020498630634055, + "instruction_rms": 0.0010020498627110325, + "raw_apical_rms": 0.0846343379427134, + "teaching_rms": 0.0010020498630634055, + "traffic_rms": 0.08462419668095914 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.6831539656768655e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55633557789611e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008095358785006215, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006404352074414491 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205636143684387, + "mean_feedback_forward_cosine": 0.9997495605099586, + "mean_feedback_forward_relative_error": 0.02198745480047599, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010033188583081752, + "instruction_rms": 0.0010033188583717864, + "raw_apical_rms": 0.08463495361823599, + "teaching_rms": 0.0010033188583081752, + "traffic_rms": 0.08462474463309623 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.7923979573074575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5509895400250203e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007997716418635352, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006348115804195404 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056331634521484, + "mean_feedback_forward_cosine": 0.9997495720463414, + "mean_feedback_forward_relative_error": 0.02198738265301912, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000945510766176803, + "instruction_rms": 0.000945510765997493, + "raw_apical_rms": 0.0846314399115287, + "teaching_rms": 0.000945510766176803, + "traffic_rms": 0.08462238563513821 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5248089841056753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5581877691237258e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008076953528522383, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006077141940742731 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0320562981069088, + "mean_feedback_forward_cosine": 0.999749570123611, + "mean_feedback_forward_relative_error": 0.02198730855278911, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00102458379317182, + "instruction_rms": 0.0010245837930899322, + "raw_apical_rms": 0.08463160274168846, + "teaching_rms": 0.00102458379317182, + "traffic_rms": 0.08462102132901504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.151085437025358e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555968560466788e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00809513667317638, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006415005577653647 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205627202987671, + "mean_feedback_forward_cosine": 0.9997495585872281, + "mean_feedback_forward_relative_error": 0.021987238057678746, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009684690597639544, + "instruction_rms": 0.0009684690596453609, + "raw_apical_rms": 0.08462028090699916, + "teaching_rms": 0.0009684690597639544, + "traffic_rms": 0.08461079467241205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.265353143433458e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5567662536602234e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008102410627329417, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006293740921169519 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205624222755432, + "mean_feedback_forward_cosine": 0.9997495374371929, + "mean_feedback_forward_relative_error": 0.021987164798643318, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009755682158272548, + "instruction_rms": 0.000975568215529958, + "raw_apical_rms": 0.08461512225964067, + "teaching_rms": 0.0009755682158272548, + "traffic_rms": 0.08460528313247594 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339475095272064, + "maximum_absolute_post_projection_soma_slope": 2.8293257514633297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558493241975701e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008027609729193294, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006420630967915058 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205621987581253, + "mean_feedback_forward_cosine": 0.99974956627815, + "mean_feedback_forward_relative_error": 0.021987092185525164, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009939295355575488, + "instruction_rms": 0.0009939295358582666, + "raw_apical_rms": 0.08461828520534666, + "teaching_rms": 0.0009939295355575488, + "traffic_rms": 0.0846086587462049 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8858607947768178e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552516115559172e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008002757362640616, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006221403937786817 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205619007349014, + "mean_feedback_forward_cosine": 0.9997495624326891, + "mean_feedback_forward_relative_error": 0.021987019557385676, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010022463724885253, + "instruction_rms": 0.0010022463725116845, + "raw_apical_rms": 0.08461402826079983, + "teaching_rms": 0.0010022463724885253, + "traffic_rms": 0.08460395405772864 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.8469171908795943e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5567877869529386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008020365758867129, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006556921216435731 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056160271167755, + "mean_feedback_forward_cosine": 0.9997495528190367, + "mean_feedback_forward_relative_error": 0.021986948221080726, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010093778616662933, + "instruction_rms": 0.0010093778613708784, + "raw_apical_rms": 0.0846190872794773, + "teaching_rms": 0.0010093778616662933, + "traffic_rms": 0.08460841015676575 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 3.505197554432016e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5583391583711716e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008014049761277826, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0063870408262312415 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032056137919425964, + "mean_feedback_forward_cosine": 0.9997495682008805, + "mean_feedback_forward_relative_error": 0.021986879708786166, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00096379689194706, + "instruction_rms": 0.0009637968916893902, + "raw_apical_rms": 0.08461680450637689, + "teaching_rms": 0.00096379689194706, + "traffic_rms": 0.08460750582674122 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9052934280571208e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5572649672196407e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008139760296501626, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006208665335252881 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205610439181328, + "mean_feedback_forward_cosine": 0.9997495720463414, + "mean_feedback_forward_relative_error": 0.021986806494814735, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010049983975515442, + "instruction_rms": 0.0010049983975929705, + "raw_apical_rms": 0.08461766990294084, + "teaching_rms": 0.0010049983975515442, + "traffic_rms": 0.08460719932424446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.7040554684276685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556402926755914e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00803023091736073, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006334791790693998 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205607458949089, + "mean_feedback_forward_cosine": 0.9997495605099586, + "mean_feedback_forward_relative_error": 0.02198673359629127, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009943122122782109, + "instruction_rms": 0.0009943122121915979, + "raw_apical_rms": 0.08461850004618557, + "teaching_rms": 0.0009943122122782109, + "traffic_rms": 0.08460779044643685 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.9584640515167848e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555625464472957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008063228466963727, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.006631418948993086 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0320560485124588, + "mean_feedback_forward_cosine": 0.9997495605099586, + "mean_feedback_forward_relative_error": 0.02198666095313045, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010201716745599865, + "instruction_rms": 0.001020171674360803, + "raw_apical_rms": 0.08461917577619991, + "teaching_rms": 0.0010201716745599865, + "traffic_rms": 0.08460831555889152 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.3712544816589798e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5576616048528345e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008098960361323994, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006379286336749792 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205602988600731, + "mean_feedback_forward_cosine": 0.9997495816599938, + "mean_feedback_forward_relative_error": 0.02198659454382235, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009182792900501016, + "instruction_rms": 0.0009182792903250644, + "raw_apical_rms": 0.08461888804601472, + "teaching_rms": 0.0009182792900501016, + "traffic_rms": 0.08461040556161319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.548472188834694e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5566185028608155e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008097171684769202, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005835805299319327 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032055992633104324, + "mean_feedback_forward_cosine": 0.9997495643554195, + "mean_feedback_forward_relative_error": 0.0219865235379867, + "min_feedback_forward_cosine": 0.9994877576828003 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009214239525856193, + "instruction_rms": 0.0009214239527428856, + "raw_apical_rms": 0.08461644382119267, + "teaching_rms": 0.0009214239525856193, + "traffic_rms": 0.08460766022249996 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.936252840868292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55375899470512e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007992713314014355, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.00579205192297697 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032055970281362534, + "mean_feedback_forward_cosine": 0.9997495855054548, + "mean_feedback_forward_relative_error": 0.021986449888396646, + "min_feedback_forward_cosine": 0.9994878768920898 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009953991815410635, + "instruction_rms": 0.0009953991816785803, + "raw_apical_rms": 0.08461607220280493, + "teaching_rms": 0.0009953991815410635, + "traffic_rms": 0.08460633481017359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5814562931714136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.554607755227586e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008140653412696292, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0062616572424955665 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205593675374985, + "mean_feedback_forward_cosine": 0.9997495758918024, + "mean_feedback_forward_relative_error": 0.02198637978384091, + "min_feedback_forward_cosine": 0.9994878172874451 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009330463627356266, + "instruction_rms": 0.0009330463626894913, + "raw_apical_rms": 0.08461425118418563, + "teaching_rms": 0.0009330463627356266, + "traffic_rms": 0.08460537313382223 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5045620688501913e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556403009917614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00818392018305014, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005891993966549635 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03205590322613716, + "mean_feedback_forward_cosine": 0.9997495874281852, + "mean_feedback_forward_relative_error": 0.02198631041533043, + "min_feedback_forward_cosine": 0.9994879961013794 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009432052631911234, + "instruction_rms": 0.0009432052632129908, + "raw_apical_rms": 0.08461482922212817, + "teaching_rms": 0.0009432052631911234, + "traffic_rms": 0.08460563191315822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.08339474350214005, + "maximum_absolute_post_projection_soma_slope": 2.5194417219154275e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5541541775179087e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008058275272734602, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005891606057062745 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9229, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3967433898925781, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2529336320, + "peak_memory_reserved_bytes": 3821010944, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 4.687810672976411e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 4.687810672976411e-08, + "min_residual_soma_slope": -4.520915908301504e-08, + "mse": 4.727982913366485e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.226943794806031e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.226943794806031e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1475834846496582, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.04273486137390137, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 14858.050523757935, + "train_wall_s": 14854.18590760231 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.02175460383296013, + 0.013336348347365856, + 0.009783044457435608, + 0.00898885726928711, + 0.005923474207520485, + 0.006553402170538902, + 0.0038889478892087936, + 0.004754113964736462, + 0.002926577115431428, + 0.0037760213017463684, + 0.002342476276680827, + 0.005092362407594919, + 0.0035116823855787516, + 0.004277744796127081, + 0.0028089198749512434, + 0.0035794125869870186, + 0.0023919157683849335, + 0.003378334455192089, + 0.002024646382778883, + 0.0026940391398966312, + 0.0017467448487877846, + 0.003947919234633446, + 0.0027886261232197285, + 0.003526116255670786, + 0.0024985498748719692, + 0.0032006092369556427, + 0.0022499680053442717, + 0.0029331750702112913, + 0.0021160657051950693, + 0.0027568177320063114, + 0.0020793545991182327 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.0, + 3.999999761581421, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 3.999999761581421, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.11566921323537827, + 0.07079319655895233, + 0.0348849780857563, + 0.048351991921663284, + 0.016349442303180695, + 0.036473244428634644, + 0.009005839936435223, + 0.025443613529205322, + 0.006116918288171291, + 0.020323241129517555, + 0.00427129166200757, + 0.02655886486172676, + 0.007858424447476864, + 0.022830985486507416, + 0.005552944727241993, + 0.018359895795583725, + 0.004318201914429665, + 0.018684223294258118, + 0.0033378105144947767, + 0.014095356687903404, + 0.002663316437974572, + 0.020948903635144234, + 0.00544493505731225, + 0.018528027459979057, + 0.00444211158901453, + 0.017210181802511215, + 0.003724918933585286, + 0.015407051891088486, + 0.003246729262173176, + 0.014654136262834072, + 0.002993625123053789 + ], + "unscaled_traffic_rms": [ + 0.7523040175437927, + 0.7535384297370911, + 1.121748685836792, + 0.743618369102478, + 1.449217438697815, + 0.7187078595161438, + 1.7273005247116089, + 0.7473960518836975, + 1.9137591123580933, + 0.743192732334137, + 2.1936936378479004, + 0.7669548392295837, + 1.7874739170074463, + 0.7494630217552185, + 2.0233731269836426, + 0.7798328995704651, + 2.2156591415405273, + 0.7232486009597778, + 2.426316738128662, + 0.7645182013511658, + 2.6234130859375, + 0.7538187503814697, + 2.0486018657684326, + 0.7612502574920654, + 2.249875783920288, + 0.7438873648643494, + 2.416125535964966, + 0.7615149617195129, + 2.6070122718811035, + 0.7525022625923157, + 2.778376817703247 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684206618894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 72745580756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 2745665106100224 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json new file mode 100644 index 0000000..bcec5ba --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json @@ -0,0 +1,6861 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5011, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9995903253555298, + "feedback_forward_cosine": [ + 0.9998393058776855, + 0.9998425841331482, + 0.9997931122779846, + 0.9997373223304749, + 0.9997537136077881, + 0.9996713399887085, + 0.9996196031570435, + 0.9995112419128418, + 0.9998084306716919, + 0.9997016191482544, + 0.9997225403785706, + 0.9998138546943665, + 0.999776303768158, + 0.9997304677963257, + 0.9997777938842773, + 0.9997199773788452, + 0.9997802376747131, + 0.9997098445892334, + 0.9997881650924683, + 0.9996970295906067, + 0.9997903108596802, + 0.9998658299446106, + 0.9998389482498169, + 0.9998205900192261, + 0.9998304843902588, + 0.999771237373352, + 0.9997951984405518, + 0.999674916267395, + 0.9997724294662476, + 0.9995083808898926, + 0.9999850392341614 + ], + "feedback_forward_norm_ratio": [ + 0.9998177289962769, + 1.0005102157592773, + 0.9998642802238464, + 1.0012266635894775, + 0.999540388584137, + 1.0020116567611694, + 0.9995760321617126, + 1.002004623413086, + 0.9987740516662598, + 1.0021671056747437, + 0.998420774936676, + 1.0004249811172485, + 0.9994477033615112, + 1.001162052154541, + 0.9992409944534302, + 1.0009486675262451, + 0.9995359182357788, + 1.0016334056854248, + 0.9993464946746826, + 1.001138687133789, + 0.9992688298225403, + 1.0005548000335693, + 0.9995976686477661, + 1.0004544258117676, + 0.9995250105857849, + 1.0010119676589966, + 0.9996137022972107, + 1.001089334487915, + 0.9996390342712402, + 1.001050353050232, + 1.0027774572372437 + ], + "feedback_forward_relative_error": [ + 0.017924342304468155, + 0.017754344269633293, + 0.020338894799351692, + 0.022967584431171417, + 0.022193925455212593, + 0.025744901970028877, + 0.02758357860147953, + 0.03135954588651657, + 0.01960141770541668, + 0.024551130831241608, + 0.023588959127664566, + 0.019300924614071846, + 0.021157586947083473, + 0.02326112799346447, + 0.021080080419778824, + 0.023696647956967354, + 0.020960435271263123, + 0.024163294583559036, + 0.020589381456375122, + 0.024656960740685463, + 0.02048652432858944, + 0.01639280840754509, + 0.017950978130102158, + 0.01895267143845558, + 0.01841617189347744, + 0.0214255228638649, + 0.020239848643541336, + 0.02553929202258587, + 0.021332532167434692, + 0.031390681862831116, + 0.006141501944512129 + ], + "innovation_negative_gradient_cosine": [ + 0.9996150732040405, + 0.9995710253715515, + 0.9996081590652466, + 0.9995531439781189, + 0.9996228218078613, + 0.9995498657226562, + 0.9996355772018433, + 0.9994957447052002, + 0.99965500831604, + 0.9995968341827393, + 0.9996651411056519, + 0.9996426105499268, + 0.999691367149353, + 0.9996111989021301, + 0.9997058510780334, + 0.9996367692947388, + 0.9997190833091736, + 0.9996544122695923, + 0.999728798866272, + 0.9996733069419861, + 0.9997476935386658, + 0.9997837543487549, + 0.9997960329055786, + 0.9997931718826294, + 0.99980628490448, + 0.9998067617416382, + 0.9998337626457214, + 0.9998568892478943, + 0.999924898147583, + 0.9999518394470215, + 0.7934749126434326 + ], + "instruction_negative_gradient_cosine": [ + 0.9996151328086853, + 0.9995710849761963, + 0.9996081590652466, + 0.9995530843734741, + 0.9996228218078613, + 0.9995498657226562, + 0.999635636806488, + 0.9994957447052002, + 0.99965500831604, + 0.9995968341827393, + 0.9996651411056519, + 0.9996426105499268, + 0.999691367149353, + 0.9996111989021301, + 0.9997059106826782, + 0.9996367692947388, + 0.9997190833091736, + 0.9996544122695923, + 0.999728798866272, + 0.9996733069419861, + 0.999747633934021, + 0.9997837543487549, + 0.9997960329055786, + 0.9997931718826294, + 0.99980628490448, + 0.9998067617416382, + 0.9998337626457214, + 0.9998568892478943, + 0.999924898147583, + 0.9999518394470215, + 0.8024725317955017 + ], + "matched_negative_gradient_cosine": [ + -0.006497480440884829, + -0.0010007985401898623, + -0.004768798593431711, + 0.0029562427662312984, + 0.0028078125324100256, + 0.004375014454126358, + 0.009340014308691025, + 0.005763593129813671, + 0.010660374537110329, + 0.010545456781983376, + 0.01165374182164669, + 0.04071634262800217, + 0.01207689382135868, + 0.04588733986020088, + -0.00496321264654398, + 0.027296870946884155, + 0.0014145260211080313, + 0.044059060513973236, + -0.0026559161487966776, + 0.01349729485809803, + -0.0007202508859336376, + 0.020670048892498016, + 0.009963254444301128, + 0.014850533567368984, + 0.06023435667157173, + 0.12118001282215118, + 0.04876578599214554, + 0.10186578333377838, + 0.05108996480703354, + 0.046388640999794006, + 0.077696792781353 + ], + "max_norm_match_direction_error": 0.02065885066986084, + "max_norm_match_relative_error": 1.4674257897695497e-07, + "mean_feedback_forward_cosine": 0.9997563823576896, + "mean_feedback_forward_relative_error": 0.021636890292528174, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.11225628852844238, + "max_absolute_post_projection_soma_slope": 7.925833322985909e-09, + "max_absolute_pre_projection_soma_slope": 0.11225628852844238, + "max_positive_post_projection_soma_slope": 7.925833322985909e-09, + "min_post_projection_soma_slope": -6.912399097558364e-09, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4260500335837172e-08, + "pre_projection_traffic_rms_ratio": 0.009860032162847598 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.00986003216282541, + "raw_negative_gradient_cosine": [ + -0.006497480440884829, + -0.0010007983073592186, + -0.004768798127770424, + 0.002956242300570011, + 0.0028078125324100256, + 0.004375015385448933, + 0.0093400152400136, + 0.00576359499245882, + 0.010660374537110329, + 0.010545456781983376, + 0.01165374368429184, + 0.04071634262800217, + 0.01207689382135868, + 0.04588734358549118, + -0.00496321264654398, + 0.027296870946884155, + 0.0014145260211080313, + 0.04405905678868294, + -0.0026559149846434593, + 0.01349729672074318, + -0.0007202494889497757, + 0.020670050755143166, + 0.009963255375623703, + 0.014850534498691559, + 0.06023435294628143, + 0.12118001282215118, + 0.04876578971743584, + 0.10186578333377838, + 0.05108996480703354, + 0.0463886484503746, + 0.07769609987735748 + ], + "teaching_negative_gradient_cosine": [ + 0.9996150732040405, + 0.9995710253715515, + 0.9996081590652466, + 0.9995531439781189, + 0.9996228218078613, + 0.9995498657226562, + 0.9996355772018433, + 0.9994957447052002, + 0.99965500831604, + 0.9995968341827393, + 0.9996651411056519, + 0.9996426105499268, + 0.999691367149353, + 0.9996111989021301, + 0.9997058510780334, + 0.9996367692947388, + 0.9997190833091736, + 0.9996544122695923, + 0.999728798866272, + 0.9996733069419861, + 0.9997476935386658, + 0.9997837543487549, + 0.9997960329055786, + 0.9997931718826294, + 0.99980628490448, + 0.9998067617416382, + 0.9998337626457214, + 0.9998568892478943, + 0.999924898147583, + 0.9999518394470215, + 0.7934749126434326 + ], + "traffic_instruction_rms_ratio": 119.9707903135327, + "used_negative_gradient_cosine": [ + 0.9996150732040405, + 0.9995710253715515, + 0.9996081590652466, + 0.9995531439781189, + 0.9996228218078613, + 0.9995498657226562, + 0.9996355772018433, + 0.9994957447052002, + 0.99965500831604, + 0.9995968341827393, + 0.9996651411056519, + 0.9996426105499268, + 0.999691367149353, + 0.9996111989021301, + 0.9997058510780334, + 0.9996367692947388, + 0.9997190833091736, + 0.9996544122695923, + 0.999728798866272, + 0.9996733069419861, + 0.9997476935386658, + 0.9997837543487549, + 0.9997960329055786, + 0.9997931718826294, + 0.99980628490448, + 0.9998067617416382, + 0.9998337626457214, + 0.9998568892478943, + 0.999924898147583, + 0.9999518394470215, + 0.7934749126434326 + ], + "wall_s": 0.1894664764404297 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3496880531311035, + "mean_feedback_forward_cosine": 0.23779246114915417, + "mean_feedback_forward_relative_error": 1.234508550936176, + "min_feedback_forward_cosine": 0.09770594537258148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.001792261825803338, + "instruction_rms": 0.001792261826251293, + "raw_apical_rms": 0.028805227960451953, + "teaching_rms": 0.001792261825803338, + "traffic_rms": 0.028727260303247956 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225633323192596, + "maximum_absolute_post_projection_soma_slope": 5.258212354419811e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.045556571320131e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010801846595000409, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.883175189743042 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3214671611785889, + "mean_feedback_forward_cosine": 0.33310096013930535, + "mean_feedback_forward_relative_error": 1.1533651851838636, + "min_feedback_forward_cosine": 0.13528896868228912 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002421221190249264, + "instruction_rms": 0.0024212211899576033, + "raw_apical_rms": 0.034849637741896075, + "teaching_rms": 0.002421221190249264, + "traffic_rms": 0.03476434366891238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 2.841998281155611e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7825343562887417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011862241955242688, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.317690772628784 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.286177396774292, + "mean_feedback_forward_cosine": 0.4128204747553795, + "mean_feedback_forward_relative_error": 1.0815526227797232, + "min_feedback_forward_cosine": 0.1772659718990326 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002908956055007506, + "instruction_rms": 0.0029089560558632085, + "raw_apical_rms": 0.03979513620461615, + "teaching_rms": 0.002908956055007506, + "traffic_rms": 0.039687916265938566 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 2.933276199712509e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7591173112030292e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009998382364186997, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0670323083877564 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2463759183883667, + "mean_feedback_forward_cosine": 0.47859098209488776, + "mean_feedback_forward_relative_error": 1.019184765315825, + "min_feedback_forward_cosine": 0.22580194473266602 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003136969448048461, + "instruction_rms": 0.0031369694475944293, + "raw_apical_rms": 0.04359462530079327, + "teaching_rms": 0.003136969448048461, + "traffic_rms": 0.043480824283611615 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.594250503307194e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7148236461990264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010957993229569451, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9066353808403015 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2118828296661377, + "mean_feedback_forward_cosine": 0.5330047703558399, + "mean_feedback_forward_relative_error": 0.9634942104739528, + "min_feedback_forward_cosine": 0.2751264274120331 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033078185183274328, + "instruction_rms": 0.003307818519762654, + "raw_apical_rms": 0.04585935586837421, + "teaching_rms": 0.0033078185183274328, + "traffic_rms": 0.04573897000460993 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 2.9385491373545847e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6795564678856313e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011519401571881556, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7787144205856323 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1795347929000854, + "mean_feedback_forward_cosine": 0.5791930998525312, + "mean_feedback_forward_relative_error": 0.9135592627909875, + "min_feedback_forward_cosine": 0.31856682896614075 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034067994728327, + "instruction_rms": 0.0034067994719840793, + "raw_apical_rms": 0.04818388201234819, + "teaching_rms": 0.0034067994728327, + "traffic_rms": 0.04806230353498617 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.8454007150457983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6583452429032212e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01264221682082263, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6978113789558411 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.149438738822937, + "mean_feedback_forward_cosine": 0.6168919176824631, + "mean_feedback_forward_relative_error": 0.8710372524876748, + "min_feedback_forward_cosine": 0.3552587628364563 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00336719592465616, + "instruction_rms": 0.0033671959250112307, + "raw_apical_rms": 0.04975851425289233, + "teaching_rms": 0.00336719592465616, + "traffic_rms": 0.04964341781125968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.2966526336840616e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351101400811515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013324666793379182, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6338754903030396 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1214444637298584, + "mean_feedback_forward_cosine": 0.6504959223731872, + "mean_feedback_forward_relative_error": 0.8313341275338204, + "min_feedback_forward_cosine": 0.38811126351356506 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003319322473813083, + "instruction_rms": 0.0033193224743269328, + "raw_apical_rms": 0.051181383486642563, + "teaching_rms": 0.003319322473813083, + "traffic_rms": 0.05107260821767357 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.5319559782133183e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6337099496498826e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013492020812763473, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5945274598693847 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0947422981262207, + "mean_feedback_forward_cosine": 0.6788346402106746, + "mean_feedback_forward_relative_error": 0.7962960206693218, + "min_feedback_forward_cosine": 0.41831809282302856 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003329283437906273, + "instruction_rms": 0.0033292834377449936, + "raw_apical_rms": 0.05222263351215701, + "teaching_rms": 0.003329283437906273, + "traffic_rms": 0.05211526161502879 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.328970438725264e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6177858545709823e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013521616016321129, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5460585849380493 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0678207874298096, + "mean_feedback_forward_cosine": 0.7045119395179134, + "mean_feedback_forward_relative_error": 0.7631245707311938, + "min_feedback_forward_cosine": 0.44739997386932373 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033346312875229315, + "instruction_rms": 0.0033346312873691673, + "raw_apical_rms": 0.05295629358292148, + "teaching_rms": 0.0033346312875229315, + "traffic_rms": 0.05285022357265569 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.1081618228890875e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.61057395972827e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013158838708820633, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5156999127006531 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0406087636947632, + "mean_feedback_forward_cosine": 0.7277736356181483, + "mean_feedback_forward_relative_error": 0.7318284290452157, + "min_feedback_forward_cosine": 0.4766763150691986 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033344675920963327, + "instruction_rms": 0.003334467591046028, + "raw_apical_rms": 0.05374417999136594, + "teaching_rms": 0.0033344675920963327, + "traffic_rms": 0.053639381114595035 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.8007780983330122e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.612133514462808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01373393522368179, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.49187483179092406 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0131824016571045, + "mean_feedback_forward_cosine": 0.7490967935131442, + "mean_feedback_forward_relative_error": 0.7020098365122273, + "min_feedback_forward_cosine": 0.504758358001709 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033325480493578783, + "instruction_rms": 0.0033325480497094274, + "raw_apical_rms": 0.0546010520836132, + "teaching_rms": 0.0033325480493578783, + "traffic_rms": 0.05449803404631906 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.974685919809872e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.602862778984721e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01347280341987151, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4723786288070679 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9852772951126099, + "mean_feedback_forward_cosine": 0.7686242422749919, + "mean_feedback_forward_relative_error": 0.6736344320158805, + "min_feedback_forward_cosine": 0.5321928262710571 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033702792408124055, + "instruction_rms": 0.003370279240280608, + "raw_apical_rms": 0.0554730912311837, + "teaching_rms": 0.0033702792408124055, + "traffic_rms": 0.055369326061221094 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.5364050415532802e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591794573869333e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012672975581042403, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.45233731906890867 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9574909806251526, + "mean_feedback_forward_cosine": 0.7865343324599727, + "mean_feedback_forward_relative_error": 0.6466560897327238, + "min_feedback_forward_cosine": 0.5584561228752136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033300708562622908, + "instruction_rms": 0.0033300708571274216, + "raw_apical_rms": 0.05599120307503897, + "teaching_rms": 0.0033300708562622908, + "traffic_rms": 0.05589091450250531 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.8411470509581704e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.594007736784222e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01258746363288867, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4323750767326355 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9284768104553223, + "mean_feedback_forward_cosine": 0.802930360840213, + "mean_feedback_forward_relative_error": 0.6208677892723391, + "min_feedback_forward_cosine": 0.5856777429580688 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033243186649561823, + "instruction_rms": 0.0033243186649019175, + "raw_apical_rms": 0.056545813755457225, + "teaching_rms": 0.0033243186649561823, + "traffic_rms": 0.05644673637626801 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.5371388545636364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591289395230428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012874887660320094, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.41533927255630493 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9013845920562744, + "mean_feedback_forward_cosine": 0.8178992521378302, + "mean_feedback_forward_relative_error": 0.5966463574478703, + "min_feedback_forward_cosine": 0.6095059514045715 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033253388544950877, + "instruction_rms": 0.0033253388545099577, + "raw_apical_rms": 0.057039058299791084, + "teaching_rms": 0.0033253388544950877, + "traffic_rms": 0.05694071077060974 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 2.6628805827044744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5850776735710754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013231701546586847, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39967508281707764 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.874859631061554, + "mean_feedback_forward_cosine": 0.831680578570212, + "mean_feedback_forward_relative_error": 0.5733389229543747, + "min_feedback_forward_cosine": 0.6320400238037109 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033104020801533684, + "instruction_rms": 0.0033104020802953668, + "raw_apical_rms": 0.05738609073486056, + "teaching_rms": 0.0033104020801533684, + "traffic_rms": 0.057289031942430374 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 2.599791493196335e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.590709769113238e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012404134778190896, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3828320609664917 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8475277423858643, + "mean_feedback_forward_cosine": 0.8443116776404842, + "mean_feedback_forward_relative_error": 0.5510564667563285, + "min_feedback_forward_cosine": 0.654889702796936 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032769496810639426, + "instruction_rms": 0.003276949680155566, + "raw_apical_rms": 0.057786097594853966, + "teaching_rms": 0.0032769496810639426, + "traffic_rms": 0.057691604400210664 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.497487372465912e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5849717691134332e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012076421928275596, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.37322906542778017 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8215506672859192, + "mean_feedback_forward_cosine": 0.8562665158702482, + "mean_feedback_forward_relative_error": 0.5293006781608828, + "min_feedback_forward_cosine": 0.6758486032485962 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033048215272533945, + "instruction_rms": 0.003304821526720537, + "raw_apical_rms": 0.05803315554951797, + "teaching_rms": 0.0033048215272533945, + "traffic_rms": 0.05793751029686721 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.042416096832312e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5808096129596566e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01197672741514764, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.36867270792007445 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7943539023399353, + "mean_feedback_forward_cosine": 0.867316807470014, + "mean_feedback_forward_relative_error": 0.508296672855654, + "min_feedback_forward_cosine": 0.6967917680740356 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033173971083641377, + "instruction_rms": 0.0033173971078284755, + "raw_apical_rms": 0.05854039350209792, + "teaching_rms": 0.0033173971083641377, + "traffic_rms": 0.058445061830587126 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.911373186975652e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.576937532019565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012154895988473236, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3556933372116089 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7677993178367615, + "mean_feedback_forward_cosine": 0.8773702729132867, + "mean_feedback_forward_relative_error": 0.48845543832548205, + "min_feedback_forward_cosine": 0.7169189453125 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033073317765581547, + "instruction_rms": 0.0033073317772349823, + "raw_apical_rms": 0.058737001971614314, + "teaching_rms": 0.0033073317765581547, + "traffic_rms": 0.05864225026309258 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.6256198992768986e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5766312767555807e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011327904578062476, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3399512421417236 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7417647838592529, + "mean_feedback_forward_cosine": 0.8868064476597693, + "mean_feedback_forward_relative_error": 0.4691789294442823, + "min_feedback_forward_cosine": 0.7359968423843384 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003302059023453005, + "instruction_rms": 0.0033020590234496476, + "raw_apical_rms": 0.059311878266022786, + "teaching_rms": 0.003302059023453005, + "traffic_rms": 0.05921828083594166 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.038580942416047e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5777768928539136e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012320597210774149, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32902210988998415 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7155755162239075, + "mean_feedback_forward_cosine": 0.8953168392181396, + "mean_feedback_forward_relative_error": 0.4510100416598781, + "min_feedback_forward_cosine": 0.7544387578964233 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032708049472139214, + "instruction_rms": 0.003270804946515275, + "raw_apical_rms": 0.05951535872565163, + "teaching_rms": 0.0032708049472139214, + "traffic_rms": 0.05942376447652585 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.084525701524399e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5754792433868905e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01261531597454441, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.32357191411972047 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6905362010002136, + "mean_feedback_forward_cosine": 0.9032265736210731, + "mean_feedback_forward_relative_error": 0.43349189383368336, + "min_feedback_forward_cosine": 0.7709805965423584 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003272499729795818, + "instruction_rms": 0.0032724997287085777, + "raw_apical_rms": 0.05997283825635708, + "teaching_rms": 0.003272499729795818, + "traffic_rms": 0.059881926658211744 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.0635067815865114e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5718286186657776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012313924905851455, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3190615742301941 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6647300720214844, + "mean_feedback_forward_cosine": 0.9107316117132863, + "mean_feedback_forward_relative_error": 0.41629005103341993, + "min_feedback_forward_cosine": 0.7878142595291138 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032740237389923374, + "instruction_rms": 0.00327402373798544, + "raw_apical_rms": 0.05992498399358805, + "teaching_rms": 0.0032740237389923374, + "traffic_rms": 0.05983376415038004 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.6139975517480707e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5709855750688462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011235115731961103, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.30554208899497987 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6416481137275696, + "mean_feedback_forward_cosine": 0.9172263453083653, + "mean_feedback_forward_relative_error": 0.40068683605040273, + "min_feedback_forward_cosine": 0.8020126223564148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032305134027988368, + "instruction_rms": 0.0032305134013642864, + "raw_apical_rms": 0.06023654667332525, + "teaching_rms": 0.0032305134027988368, + "traffic_rms": 0.0601481441017935 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.247811264373013e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5675895160046115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010335527925943647, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2948358810806274 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6188873648643494, + "mean_feedback_forward_cosine": 0.9233948857553543, + "mean_feedback_forward_relative_error": 0.38522597378300083, + "min_feedback_forward_cosine": 0.815637469291687 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032253577353462603, + "instruction_rms": 0.0032253577350284585, + "raw_apical_rms": 0.060719556235803676, + "teaching_rms": 0.0032253577353462603, + "traffic_rms": 0.06063212153176193 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.7535248747767582e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5718172491880183e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010918030632283382, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29439870826721193 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5944331288337708, + "mean_feedback_forward_cosine": 0.929389076848184, + "mean_feedback_forward_relative_error": 0.3697683407414344, + "min_feedback_forward_cosine": 0.8296734094619751 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033057042535208584, + "instruction_rms": 0.00330570425395004, + "raw_apical_rms": 0.06085092983549735, + "teaching_rms": 0.0033057042535208584, + "traffic_rms": 0.060759507640024746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.396624848051033e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5801276581406607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01133665934687235, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.29211154641151427 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5716933608055115, + "mean_feedback_forward_cosine": 0.9346478369928175, + "mean_feedback_forward_relative_error": 0.3556915093814173, + "min_feedback_forward_cosine": 0.8423337936401367 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032548743347958486, + "instruction_rms": 0.003254874335233672, + "raw_apical_rms": 0.060817232562968754, + "teaching_rms": 0.0032548743347958486, + "traffic_rms": 0.06072845435202621 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.0558943819869455e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5669451091088644e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01192258029651461, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28162732667922974 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.549853503704071, + "mean_feedback_forward_cosine": 0.9396389838187925, + "mean_feedback_forward_relative_error": 0.34180730773556617, + "min_feedback_forward_cosine": 0.8540327548980713 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003254695845098277, + "instruction_rms": 0.003254695844725547, + "raw_apical_rms": 0.061029588091336545, + "teaching_rms": 0.003254695845098277, + "traffic_rms": 0.06094101024208075 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.4466967118996763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5584039433857285e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012014019513851422, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2790811213731766 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5292419195175171, + "mean_feedback_forward_cosine": 0.9441387634123525, + "mean_feedback_forward_relative_error": 0.3288032722569281, + "min_feedback_forward_cosine": 0.8645808696746826 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032379930618421037, + "instruction_rms": 0.0032379930630742826, + "raw_apical_rms": 0.06097692898944999, + "teaching_rms": 0.0032379930618421037, + "traffic_rms": 0.06088911481139244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 3.1269017597423954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5597649515233644e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012800837947494967, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2696018798828125 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5083063840866089, + "mean_feedback_forward_cosine": 0.9483677571819674, + "mean_feedback_forward_relative_error": 0.316038433582552, + "min_feedback_forward_cosine": 0.8749896287918091 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032568659113997077, + "instruction_rms": 0.0032568659122541197, + "raw_apical_rms": 0.06110950468634659, + "teaching_rms": 0.0032568659113997077, + "traffic_rms": 0.061020943497906495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 3.495733480463059e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5610257849241372e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013039541485083609, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26743567897796633 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4887472987174988, + "mean_feedback_forward_cosine": 0.9522715191687307, + "mean_feedback_forward_relative_error": 0.30383396004476854, + "min_feedback_forward_cosine": 0.8842469453811646 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003252115028098189, + "instruction_rms": 0.0032521150274152965, + "raw_apical_rms": 0.06112757169079033, + "teaching_rms": 0.003252115028098189, + "traffic_rms": 0.06103910094756242 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.284615246457179e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5590828970195226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013239898182371835, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2594930760860443 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46973827481269836, + "mean_feedback_forward_cosine": 0.955971485184085, + "mean_feedback_forward_relative_error": 0.29178578526743, + "min_feedback_forward_cosine": 0.8929667472839355 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033002495384564625, + "instruction_rms": 0.003300249538139308, + "raw_apical_rms": 0.06152919953251131, + "teaching_rms": 0.0033002495384564625, + "traffic_rms": 0.061438837462397373 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.9531179279729258e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562754421078797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012943892439867828, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25627604156017303 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45103105902671814, + "mean_feedback_forward_cosine": 0.9593049499296373, + "mean_feedback_forward_relative_error": 0.2804588778845726, + "min_feedback_forward_cosine": 0.9011492729187012 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032603065747262845, + "instruction_rms": 0.0032603065750236126, + "raw_apical_rms": 0.06168654494614401, + "teaching_rms": 0.0032603065747262845, + "traffic_rms": 0.06159845643913828 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.823919397267673e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55961333243947e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01201314108154925, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25371675619125367 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4319956302642822, + "mean_feedback_forward_cosine": 0.9625148369419959, + "mean_feedback_forward_relative_error": 0.2691259352910903, + "min_feedback_forward_cosine": 0.909275233745575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003300444097266203, + "instruction_rms": 0.003300444096538425, + "raw_apical_rms": 0.061765732031903955, + "teaching_rms": 0.003300444097266203, + "traffic_rms": 0.06167562457036192 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.994701103238185e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5701525951142592e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013065614479541826, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25555422548294066 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41513386368751526, + "mean_feedback_forward_cosine": 0.965345157730964, + "mean_feedback_forward_relative_error": 0.25877136592903444, + "min_feedback_forward_cosine": 0.9161350727081299 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032145991802796205, + "instruction_rms": 0.0032145991796913714, + "raw_apical_rms": 0.06179143748636271, + "teaching_rms": 0.0032145991802796205, + "traffic_rms": 0.061705900928129616 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 5.264437774599173e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.561920957799572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012403182458304603, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24684164540290832 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.398198664188385, + "mean_feedback_forward_cosine": 0.9679886833313973, + "mean_feedback_forward_relative_error": 0.2486970506368145, + "min_feedback_forward_cosine": 0.9227094054222107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032494780243739587, + "instruction_rms": 0.0032494780243708804, + "raw_apical_rms": 0.06184363574251306, + "teaching_rms": 0.0032494780243739587, + "traffic_rms": 0.06175642943583635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.0539756440693964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5562731455362073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012822947046996512, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24320279186725616 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3827840983867645, + "mean_feedback_forward_cosine": 0.9703785719410065, + "mean_feedback_forward_relative_error": 0.23920422047376633, + "min_feedback_forward_cosine": 0.9285213947296143 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032236181888009954, + "instruction_rms": 0.0032236181894204218, + "raw_apical_rms": 0.06201272401368562, + "teaching_rms": 0.0032236181888009954, + "traffic_rms": 0.06192682060653752 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.947093401597158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5617415936299073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012266634909092832, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23518570545196532 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36789071559906006, + "mean_feedback_forward_cosine": 0.9726357286976229, + "mean_feedback_forward_relative_error": 0.2298748591734517, + "min_feedback_forward_cosine": 0.9339315891265869 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003259428996929, + "instruction_rms": 0.0032594289970056915, + "raw_apical_rms": 0.06241495298691332, + "teaching_rms": 0.003259428996929, + "traffic_rms": 0.06232767602460293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.644509132845087e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5593482522842794e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012249491952177081, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23642174144268036 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3538934588432312, + "mean_feedback_forward_cosine": 0.974668687389743, + "mean_feedback_forward_relative_error": 0.22114848609893553, + "min_feedback_forward_cosine": 0.9388415813446045 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032209917954261922, + "instruction_rms": 0.0032209917960821476, + "raw_apical_rms": 0.06254188649246761, + "teaching_rms": 0.0032209917954261922, + "traffic_rms": 0.06245687771193772 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.347638693502631e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5568513603443873e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012249291725254977, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22949023630142212 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3407110869884491, + "mean_feedback_forward_cosine": 0.9765158853223247, + "mean_feedback_forward_relative_error": 0.2128911619224856, + "min_feedback_forward_cosine": 0.9432579278945923 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031841620264023316, + "instruction_rms": 0.0031841620261899664, + "raw_apical_rms": 0.06263913556287237, + "teaching_rms": 0.0031841620264023316, + "traffic_rms": 0.06255632475136726 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.658540137507771e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5628003755162997e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012383326647550437, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22707825716018676 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3270176947116852, + "mean_feedback_forward_cosine": 0.9782925344282581, + "mean_feedback_forward_relative_error": 0.20464389410711104, + "min_feedback_forward_cosine": 0.9476891160011292 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003236392452746968, + "instruction_rms": 0.003236392452344772, + "raw_apical_rms": 0.06271065835380563, + "teaching_rms": 0.003236392452746968, + "traffic_rms": 0.06262489019660311 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.010047405245132e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5618930406462716e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012588131750449472, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22609810742139816 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3139725923538208, + "mean_feedback_forward_cosine": 0.9799425621186534, + "mean_feedback_forward_relative_error": 0.19673132487843115, + "min_feedback_forward_cosine": 0.9517279863357544 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032139282340533336, + "instruction_rms": 0.003213928234158405, + "raw_apical_rms": 0.06284613678725086, + "teaching_rms": 0.0032139282340533336, + "traffic_rms": 0.06276156598699817 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.4636117024765554e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5517970244312645e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012598583491059887, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22567077231884003 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30098485946655273, + "mean_feedback_forward_cosine": 0.9814900640518435, + "mean_feedback_forward_relative_error": 0.1889460392055973, + "min_feedback_forward_cosine": 0.9555969834327698 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032631915978411515, + "instruction_rms": 0.00326319159747628, + "raw_apical_rms": 0.06289389242370472, + "teaching_rms": 0.0032631915978411515, + "traffic_rms": 0.06280691631414138 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 6.71051907374931e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5551895066815612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013183087705210402, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2225477651643753 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2885339558124542, + "mean_feedback_forward_cosine": 0.9828755701741865, + "mean_feedback_forward_relative_error": 0.1817206326992281, + "min_feedback_forward_cosine": 0.9591548442840576 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032146510133345537, + "instruction_rms": 0.0032146510127874965, + "raw_apical_rms": 0.06318518193205687, + "teaching_rms": 0.0032146510133345537, + "traffic_rms": 0.0631010560451943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 5.160919869240388e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.557698739479417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013286316370171232, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.22248741296291352 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2767795920372009, + "mean_feedback_forward_cosine": 0.98417708950658, + "mean_feedback_forward_relative_error": 0.17469270707618806, + "min_feedback_forward_cosine": 0.9623985290527344 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003223346443648342, + "instruction_rms": 0.003223346443817166, + "raw_apical_rms": 0.06319716196049695, + "teaching_rms": 0.003223346443648342, + "traffic_rms": 0.06311259041357734 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.2019306195916215e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5570246728026184e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012517031915070542, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21865181411743165 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26563000679016113, + "mean_feedback_forward_cosine": 0.9853603436100867, + "mean_feedback_forward_relative_error": 0.16801612764116255, + "min_feedback_forward_cosine": 0.9653445482254028 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032019115455290265, + "instruction_rms": 0.0032019115449666634, + "raw_apical_rms": 0.0634449066605615, + "teaching_rms": 0.0032019115455290265, + "traffic_rms": 0.06336201929111465 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.898474159631405e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5558445619538108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01311208447202521, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2144723404788971 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25532397627830505, + "mean_feedback_forward_cosine": 0.9864642197085965, + "mean_feedback_forward_relative_error": 0.1615519479157463, + "min_feedback_forward_cosine": 0.9679573774337769 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032507335967120023, + "instruction_rms": 0.0032507335954448063, + "raw_apical_rms": 0.06359034115045026, + "teaching_rms": 0.0032507335967120023, + "traffic_rms": 0.06350483665831345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.883069593120126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5624518136785833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01296623457293837, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2116580925321579 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24493002891540527, + "mean_feedback_forward_cosine": 0.9874960780143738, + "mean_feedback_forward_relative_error": 0.15528010576963425, + "min_feedback_forward_cosine": 0.9704850912094116 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032284191180477328, + "instruction_rms": 0.0032284191176412385, + "raw_apical_rms": 0.06366011916848571, + "teaching_rms": 0.0032284191180477328, + "traffic_rms": 0.06357610938960903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.807570791991566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5557410757543787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01171205688162901, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.21239043782234193 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2350618988275528, + "mean_feedback_forward_cosine": 0.9884467663303498, + "mean_feedback_forward_relative_error": 0.14924997836351395, + "min_feedback_forward_cosine": 0.9727891683578491 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032528531822477976, + "instruction_rms": 0.003252853182494769, + "raw_apical_rms": 0.06399855883714173, + "teaching_rms": 0.0032528531822477976, + "traffic_rms": 0.06391358028661491 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.6961119437582965e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.559745537056108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012536972272347374, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20796265053749086 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22559034824371338, + "mean_feedback_forward_cosine": 0.9892987301272731, + "mean_feedback_forward_relative_error": 0.14361524690062769, + "min_feedback_forward_cosine": 0.9749218225479126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003208858657762247, + "instruction_rms": 0.0032088586582928753, + "raw_apical_rms": 0.06424261075082997, + "teaching_rms": 0.003208858657762247, + "traffic_rms": 0.06416018321896445 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.4010067651024656e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.549425059134784e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012514431149795651, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20420211280822753 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2162429839372635, + "mean_feedback_forward_cosine": 0.9901151426376835, + "mean_feedback_forward_relative_error": 0.1380332242096624, + "min_feedback_forward_cosine": 0.9769406318664551 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003230649462503566, + "instruction_rms": 0.0032306494628555928, + "raw_apical_rms": 0.06466205844044151, + "teaching_rms": 0.003230649462503566, + "traffic_rms": 0.06457886516143173 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.6073931397595516e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5525071400094274e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012177581630662487, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20856931941986084 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2075895220041275, + "mean_feedback_forward_cosine": 0.9908495622296487, + "mean_feedback_forward_relative_error": 0.1327750995274513, + "min_feedback_forward_cosine": 0.9787402153015137 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00323263499780507, + "instruction_rms": 0.003232634998765458, + "raw_apical_rms": 0.06484330787039168, + "teaching_rms": 0.00323263499780507, + "traffic_rms": 0.06476054526765924 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.4263422321600956e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.548854281078689e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01235526797269067, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.1985908332824707 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19919037818908691, + "mean_feedback_forward_cosine": 0.991537542112412, + "mean_feedback_forward_relative_error": 0.1276785499386249, + "min_feedback_forward_cosine": 0.9804139137268066 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032567782616892173, + "instruction_rms": 0.0032567782620157447, + "raw_apical_rms": 0.06499176034073043, + "teaching_rms": 0.0032567782616892173, + "traffic_rms": 0.06490787925330493 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.1997503192069416e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5500044898529143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011773352922101018, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.20114256310939788 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1909475475549698, + "mean_feedback_forward_cosine": 0.9921798206144764, + "mean_feedback_forward_relative_error": 0.12273224359077792, + "min_feedback_forward_cosine": 0.9819931387901306 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003214295794738556, + "instruction_rms": 0.00321429579503298, + "raw_apical_rms": 0.06498556913398877, + "teaching_rms": 0.003214295794738556, + "traffic_rms": 0.06490337558985543 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3791843040708045e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5557164605611118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012072926961486636, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.20277392561912536 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18320252001285553, + "mean_feedback_forward_cosine": 0.9927667148651615, + "mean_feedback_forward_relative_error": 0.11803499861590323, + "min_feedback_forward_cosine": 0.9834180474281311 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031965488879336116, + "instruction_rms": 0.003196548888430924, + "raw_apical_rms": 0.0652538080997281, + "teaching_rms": 0.0031965488879336116, + "traffic_rms": 0.06517323427237556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.97686754638471e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547342659498856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012198412931471204, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19817508615493776 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1758364886045456, + "mean_feedback_forward_cosine": 0.993301430056172, + "mean_feedback_forward_relative_error": 0.11360123332950377, + "min_feedback_forward_cosine": 0.9847174286842346 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00319837615440292, + "instruction_rms": 0.0031983761546225216, + "raw_apical_rms": 0.06539917326301826, + "teaching_rms": 0.00319837615440292, + "traffic_rms": 0.06531844557725296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.943426728549639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55066315256387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01110784378020459, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19506875994682313 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16860756278038025, + "mean_feedback_forward_cosine": 0.9938184061358052, + "mean_feedback_forward_relative_error": 0.10913190161508898, + "min_feedback_forward_cosine": 0.985942006111145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032437568213884225, + "instruction_rms": 0.003243756820911829, + "raw_apical_rms": 0.06556462923254196, + "teaching_rms": 0.0032437568213884225, + "traffic_rms": 0.06548210313057999 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.5059532166314966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5471049040930825e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011031462558085807, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19833106989383698 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16182535886764526, + "mean_feedback_forward_cosine": 0.994288488741844, + "mean_feedback_forward_relative_error": 0.10489301900229146, + "min_feedback_forward_cosine": 0.9870392084121704 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032350981111001565, + "instruction_rms": 0.0032350981108131937, + "raw_apical_rms": 0.06570542131652134, + "teaching_rms": 0.0032350981111001565, + "traffic_rms": 0.0656232432125218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.0090832814930764e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5511335897796128e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011353304118281886, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1921114242553711 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15514838695526123, + "mean_feedback_forward_cosine": 0.9947307244423897, + "mean_feedback_forward_relative_error": 0.10076737680262135, + "min_feedback_forward_cosine": 0.9880790710449219 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032483517716519797, + "instruction_rms": 0.0032483517715250062, + "raw_apical_rms": 0.06573850703763767, + "teaching_rms": 0.0032483517716519797, + "traffic_rms": 0.06565589789372665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.153312715970969e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5468546345603897e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011345765531539272, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19181999053001403 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14878042042255402, + "mean_feedback_forward_cosine": 0.9951383682989305, + "mean_feedback_forward_relative_error": 0.09679948730814841, + "min_feedback_forward_cosine": 0.9890398979187012 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032353399084853454, + "instruction_rms": 0.003235339907477219, + "raw_apical_rms": 0.06596452935620388, + "teaching_rms": 0.0032353399084853454, + "traffic_rms": 0.06588286260058066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.330524916440481e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5495665637289985e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011128180677743241, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19286126182079316 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14288330078125, + "mean_feedback_forward_cosine": 0.9955042350676752, + "mean_feedback_forward_relative_error": 0.09308601879785137, + "min_feedback_forward_cosine": 0.9898863434791565 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032109956478173474, + "instruction_rms": 0.003210995648243102, + "raw_apical_rms": 0.06615911595785673, + "teaching_rms": 0.0032109956478173474, + "traffic_rms": 0.06607867233061988 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.034749340713461e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550724548975256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011582143067253824, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18887831575870515 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1369892805814743, + "mean_feedback_forward_cosine": 0.9958446679576751, + "mean_feedback_forward_relative_error": 0.08949815229542794, + "min_feedback_forward_cosine": 0.990699827671051 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003194452987056155, + "instruction_rms": 0.0031944529870642567, + "raw_apical_rms": 0.06639654014049834, + "teaching_rms": 0.003194452987056155, + "traffic_rms": 0.06631740461242881 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.78869664530157e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5480334150649503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0125897769065541, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.1881254771566391 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13136789202690125, + "mean_feedback_forward_cosine": 0.9961570770509781, + "mean_feedback_forward_relative_error": 0.08607389747856124, + "min_feedback_forward_cosine": 0.9914438128471375 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032142986261198955, + "instruction_rms": 0.003214298626186256, + "raw_apical_rms": 0.0665539865606066, + "teaching_rms": 0.0032142986261198955, + "traffic_rms": 0.06647382482916069 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.52482096843687e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5514451018518957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012602388897453326, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18574215658187868 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12614275515079498, + "mean_feedback_forward_cosine": 0.9964480034766658, + "mean_feedback_forward_relative_error": 0.08275650063109014, + "min_feedback_forward_cosine": 0.9921080470085144 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003226321439153004, + "instruction_rms": 0.0032263214399045645, + "raw_apical_rms": 0.06661003530188929, + "teaching_rms": 0.003226321439153004, + "traffic_rms": 0.0665298523858992 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.858676388470485e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538883263913092e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011490582813745858, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18655368221282959 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1210818812251091, + "mean_feedback_forward_cosine": 0.9967135652419059, + "mean_feedback_forward_relative_error": 0.07959668494520648, + "min_feedback_forward_cosine": 0.9927269220352173 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003205544800714363, + "instruction_rms": 0.0032055448003530437, + "raw_apical_rms": 0.06663934777154236, + "teaching_rms": 0.003205544800714363, + "traffic_rms": 0.06656003544933559 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.765977124066012e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5507838924459323e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012904993730806174, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18691049118995667 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11632870137691498, + "mean_feedback_forward_cosine": 0.9969619897104078, + "mean_feedback_forward_relative_error": 0.07652976898656737, + "min_feedback_forward_cosine": 0.9932830333709717 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032213756391028807, + "instruction_rms": 0.0032213756392000326, + "raw_apical_rms": 0.06686097957022545, + "teaching_rms": 0.0032213756391028807, + "traffic_rms": 0.06678112165277941 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.85621853840712e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5408713788471714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012727575742147237, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18197136464118957 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1117636188864708, + "mean_feedback_forward_cosine": 0.9971885315833553, + "mean_feedback_forward_relative_error": 0.07361533733144883, + "min_feedback_forward_cosine": 0.9937996864318848 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003212438798966721, + "instruction_rms": 0.0032124387987398553, + "raw_apical_rms": 0.06706945404650862, + "teaching_rms": 0.003212438798966721, + "traffic_rms": 0.06698981599142298 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 6.094604998452269e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552361326677047e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011446031623193595, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18268239092826843 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10737975686788559, + "mean_feedback_forward_cosine": 0.9973946156040314, + "mean_feedback_forward_relative_error": 0.07086179723903056, + "min_feedback_forward_cosine": 0.9942737817764282 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003176865691925346, + "instruction_rms": 0.0031768656926891895, + "raw_apical_rms": 0.06715577213435858, + "teaching_rms": 0.003176865691925346, + "traffic_rms": 0.06707829761706467 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.0378033361321286e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550760334767396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012134748198358561, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.18076671446800233 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1030321791768074, + "mean_feedback_forward_cosine": 0.997592935639043, + "mean_feedback_forward_relative_error": 0.06812001722714593, + "min_feedback_forward_cosine": 0.9947277903556824 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032171399983352755, + "instruction_rms": 0.0032171399992673376, + "raw_apical_rms": 0.06726017080249543, + "teaching_rms": 0.0032171399983352755, + "traffic_rms": 0.06718071225364061 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 6.5172621077636e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552829266892198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013137524151754731, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18168134922981263 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09878364205360413, + "mean_feedback_forward_cosine": 0.9977741664455783, + "mean_feedback_forward_relative_error": 0.06550340985338535, + "min_feedback_forward_cosine": 0.9951515793800354 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003233248690421359, + "instruction_rms": 0.003233248689924884, + "raw_apical_rms": 0.06746123428172678, + "teaching_rms": 0.003233248690421359, + "traffic_rms": 0.0673810143742873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.7667164437825704e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5487039161640377e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01372837641302714, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1830405264377594 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09487704932689667, + "mean_feedback_forward_cosine": 0.9979389232973899, + "mean_feedback_forward_relative_error": 0.06303370617810757, + "min_feedback_forward_cosine": 0.9955255389213562 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00317904553870906, + "instruction_rms": 0.003179045538725786, + "raw_apical_rms": 0.06753658317492056, + "teaching_rms": 0.00317904553870906, + "traffic_rms": 0.06745948046760031 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.985048652703881e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373918607226653e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013177344017347008, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.17601748112678528 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09110244363546371, + "mean_feedback_forward_cosine": 0.9980943452927374, + "mean_feedback_forward_relative_error": 0.06061276613223937, + "min_feedback_forward_cosine": 0.995873749256134 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003215237442668525, + "instruction_rms": 0.003215237442885354, + "raw_apical_rms": 0.06754507153428009, + "teaching_rms": 0.003215237442668525, + "traffic_rms": 0.06746604412137065 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.231313610603138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550971673080755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012338044104587092, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17713300874233245 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0875559002161026, + "mean_feedback_forward_cosine": 0.9982354333323817, + "mean_feedback_forward_relative_error": 0.05832481750797841, + "min_feedback_forward_cosine": 0.9961892366409302 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032071822550401303, + "instruction_rms": 0.0032071822555785456, + "raw_apical_rms": 0.06765613811325459, + "teaching_rms": 0.0032071822550401303, + "traffic_rms": 0.06757779191183735 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.657900814526329e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54856537035424e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01345870207869326, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17717875088214874 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08400376886129379, + "mean_feedback_forward_cosine": 0.998368788150049, + "mean_feedback_forward_relative_error": 0.0560823836994748, + "min_feedback_forward_cosine": 0.9964916110038757 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003207843373497529, + "instruction_rms": 0.00320784337293921, + "raw_apical_rms": 0.06786469122698273, + "teaching_rms": 0.003207843373497529, + "traffic_rms": 0.06778623119068314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.143282339441612e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5522511868982402e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01251413688322687, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17683649698734283 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08047140389680862, + "mean_feedback_forward_cosine": 0.9984961767350474, + "mean_feedback_forward_relative_error": 0.05385354653962197, + "min_feedback_forward_cosine": 0.9967791438102722 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003230704648836473, + "instruction_rms": 0.003230704649396729, + "raw_apical_rms": 0.06787018661331655, + "teaching_rms": 0.003230704648836473, + "traffic_rms": 0.06779102888394425 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.6854018503618136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5564098523122494e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01281806908456249, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.18259278288841246 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07737720012664795, + "mean_feedback_forward_cosine": 0.9986069875378762, + "mean_feedback_forward_relative_error": 0.05182503842778744, + "min_feedback_forward_cosine": 0.9970215559005737 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031915509285379666, + "instruction_rms": 0.0031915509280721517, + "raw_apical_rms": 0.06786560035908847, + "teaching_rms": 0.0031915509285379666, + "traffic_rms": 0.06778774802561156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.250687706848112e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.551957394398743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013534104489478175, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.16936449822425842 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07424913346767426, + "mean_feedback_forward_cosine": 0.9987144470214844, + "mean_feedback_forward_relative_error": 0.049782115665654984, + "min_feedback_forward_cosine": 0.997256875038147 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032537916971644805, + "instruction_rms": 0.0032537916970613065, + "raw_apical_rms": 0.06796605586934941, + "teaching_rms": 0.0032537916971644805, + "traffic_rms": 0.06788521508845385 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.952806165192669e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5486913382538383e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013515807182226242, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17274986437797546 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07118213921785355, + "mean_feedback_forward_cosine": 0.9988120775068959, + "mean_feedback_forward_relative_error": 0.04785344347117409, + "min_feedback_forward_cosine": 0.9974775910377502 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032224958931585702, + "instruction_rms": 0.003222495891795626, + "raw_apical_rms": 0.06804298833463834, + "teaching_rms": 0.0032224958931585702, + "traffic_rms": 0.06796428825108079 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.583536394875409e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.548270617895912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012855574889148286, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17545533507347108 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06846463680267334, + "mean_feedback_forward_cosine": 0.9988992444930538, + "mean_feedback_forward_relative_error": 0.04606259998775298, + "min_feedback_forward_cosine": 0.9976663589477539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003199533112929753, + "instruction_rms": 0.003199533112485015, + "raw_apical_rms": 0.06828026934959136, + "teaching_rms": 0.003199533112929753, + "traffic_rms": 0.06820249427635146 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.806460213695573e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5480163523162624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012701635507546008, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16954303053855896 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06588231772184372, + "mean_feedback_forward_cosine": 0.9989807951834894, + "mean_feedback_forward_relative_error": 0.04432452752465202, + "min_feedback_forward_cosine": 0.9978392124176025 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003177771101383703, + "instruction_rms": 0.003177771102182817, + "raw_apical_rms": 0.06836480305192148, + "teaching_rms": 0.003177771101383703, + "traffic_rms": 0.06828852302405129 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.9567887572266045e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558752725948402e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012365344914549081, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.16692307882547378 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06329449266195297, + "mean_feedback_forward_cosine": 0.9990581658578688, + "mean_feedback_forward_relative_error": 0.042615234911922484, + "min_feedback_forward_cosine": 0.9980050325393677 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031972442473566947, + "instruction_rms": 0.0031972442468029, + "raw_apical_rms": 0.06868305059237105, + "teaching_rms": 0.0031972442473566947, + "traffic_rms": 0.06860572662084691 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.506414879619115e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54528325883772e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012643287029018097, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16855574934244155 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06074243783950806, + "mean_feedback_forward_cosine": 0.99912860508888, + "mean_feedback_forward_relative_error": 0.040991374772162206, + "min_feedback_forward_cosine": 0.9981624484062195 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003164836950603174, + "instruction_rms": 0.0031648369507949327, + "raw_apical_rms": 0.0691516835375497, + "teaching_rms": 0.003164836950603174, + "traffic_rms": 0.06907686707964024 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.699916189565556e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54752411881845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012617722457909503, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.17011800132751465 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05823053419589996, + "mean_feedback_forward_cosine": 0.9991959045010228, + "mean_feedback_forward_relative_error": 0.03937796584420627, + "min_feedback_forward_cosine": 0.998310923576355 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003260645049182299, + "instruction_rms": 0.0032606450486044195, + "raw_apical_rms": 0.06916980896471374, + "teaching_rms": 0.003260645049182299, + "traffic_rms": 0.06909055454973252 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.443735690529138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.559088502024985e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012758715040523097, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.17331657188177108 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05590132996439934, + "mean_feedback_forward_cosine": 0.9992569511936557, + "mean_feedback_forward_relative_error": 0.037854788915043874, + "min_feedback_forward_cosine": 0.9984432458877563 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032118374285425924, + "instruction_rms": 0.00321183742824767, + "raw_apical_rms": 0.06936774163088637, + "teaching_rms": 0.0032118374285425924, + "traffic_rms": 0.06929064361213332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 5.573931360913775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.555120771460031e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01306618340205972, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.16967018373012543 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05376424267888069, + "mean_feedback_forward_cosine": 0.9993114663708594, + "mean_feedback_forward_relative_error": 0.036435806282585664, + "min_feedback_forward_cosine": 0.9985595345497131 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031854068247147557, + "instruction_rms": 0.0031854068255197515, + "raw_apical_rms": 0.06931274481824935, + "teaching_rms": 0.0031854068247147557, + "traffic_rms": 0.06923681228095742 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.849746832429446e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5473142591393323e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01087553703058199, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16562317004203797 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05160946398973465, + "mean_feedback_forward_cosine": 0.999363322411814, + "mean_feedback_forward_relative_error": 0.03503704010959594, + "min_feedback_forward_cosine": 0.9986729621887207 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003208312977967411, + "instruction_rms": 0.003208312977837983, + "raw_apical_rms": 0.0694633649579455, + "teaching_rms": 0.003208312977967411, + "traffic_rms": 0.06938678195148312 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 2.7672133029454926e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5547131341584104e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011845395882054216, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16786659562587739 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04950564354658127, + "mean_feedback_forward_cosine": 0.9994118405926612, + "mean_feedback_forward_relative_error": 0.03367937218037344, + "min_feedback_forward_cosine": 0.998778760433197 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003216587668289991, + "instruction_rms": 0.003216587668500994, + "raw_apical_rms": 0.06977865239911522, + "teaching_rms": 0.003216587668289991, + "traffic_rms": 0.06970201198091697 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.526652475078663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5596203343676212e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01244089436491098, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16934751215457916 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.047498174011707306, + "mean_feedback_forward_cosine": 0.999456649826419, + "mean_feedback_forward_relative_error": 0.03237053896150281, + "min_feedback_forward_cosine": 0.998875617980957 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032130074334000003, + "instruction_rms": 0.0032130074324526778, + "raw_apical_rms": 0.0698778289206791, + "teaching_rms": 0.0032130074334000003, + "traffic_rms": 0.0698013212477438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.736473613320413e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.553865754120722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013044713458489782, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16322873593330384 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04556179419159889, + "mean_feedback_forward_cosine": 0.9994974194034454, + "mean_feedback_forward_relative_error": 0.031135118446282802, + "min_feedback_forward_cosine": 0.9989652633666992 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031951559741844222, + "instruction_rms": 0.0031951559744692287, + "raw_apical_rms": 0.07014216885975251, + "teaching_rms": 0.0031951559741844222, + "traffic_rms": 0.07006691361841266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.9629758080982356e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5590567431708584e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01223940719283602, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1651152680683136 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04377276450395584, + "mean_feedback_forward_cosine": 0.9995346069335938, + "mean_feedback_forward_relative_error": 0.029957545260267872, + "min_feedback_forward_cosine": 0.9990448355674744 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003146227481346474, + "instruction_rms": 0.003146227482278957, + "raw_apical_rms": 0.07049692264534389, + "teaching_rms": 0.003146227481346474, + "traffic_rms": 0.07042419909574009 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 6.381965533819312e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5452630314964514e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0124342759197001, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16343099618911744 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.041929442435503006, + "mean_feedback_forward_cosine": 0.9995713829994202, + "mean_feedback_forward_relative_error": 0.028751591550967386, + "min_feedback_forward_cosine": 0.9991235733032227 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032454267076244456, + "instruction_rms": 0.0032454267074225043, + "raw_apical_rms": 0.0704085819249939, + "teaching_rms": 0.0032454267076244456, + "traffic_rms": 0.07033109964185345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.5504549461838906e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547030511155356e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011915815871920284, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1677961159467697 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0401989109814167, + "mean_feedback_forward_cosine": 0.9996034964438407, + "mean_feedback_forward_relative_error": 0.02765348343358886, + "min_feedback_forward_cosine": 0.9991942644119263 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003182952886214682, + "instruction_rms": 0.0031829528862795287, + "raw_apical_rms": 0.07047367786412613, + "teaching_rms": 0.003182952886214682, + "traffic_rms": 0.0703989295009962 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.350704685407436e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.560535902495355e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012781502754989734, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16370555382728577 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03866253420710564, + "mean_feedback_forward_cosine": 0.999632602737796, + "mean_feedback_forward_relative_error": 0.02661593414602741, + "min_feedback_forward_cosine": 0.9992546439170837 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003150415899012918, + "instruction_rms": 0.003150415898238591, + "raw_apical_rms": 0.07041696743767233, + "teaching_rms": 0.003150415899012918, + "traffic_rms": 0.07034399953170782 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.5646153523648536e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5477677075889753e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012045216038986175, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1565270564508438 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0371161513030529, + "mean_feedback_forward_cosine": 0.9996611072171119, + "mean_feedback_forward_relative_error": 0.025566103147162546, + "min_feedback_forward_cosine": 0.999312698841095 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003217408216500059, + "instruction_rms": 0.003217408216341552, + "raw_apical_rms": 0.07052881929902027, + "teaching_rms": 0.003217408216500059, + "traffic_rms": 0.07045281206948315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.944271043110348e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5467007523041284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012766691686505604, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16316080739498137 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.035695161670446396, + "mean_feedback_forward_cosine": 0.9996861123269604, + "mean_feedback_forward_relative_error": 0.02460332614399733, + "min_feedback_forward_cosine": 0.9993643164634705 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031573797579908585, + "instruction_rms": 0.0031573797582007284, + "raw_apical_rms": 0.07054352620548839, + "teaching_rms": 0.0031573797579908585, + "traffic_rms": 0.07047023074902457 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.358387511478213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5457922815224475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011412010161796902, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1619117552089691 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03423386812210083, + "mean_feedback_forward_cosine": 0.9997104637084469, + "mean_feedback_forward_relative_error": 0.02363008774456478, + "min_feedback_forward_cosine": 0.9994153380393982 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032383314577264692, + "instruction_rms": 0.003238331458085077, + "raw_apical_rms": 0.07061522881780413, + "teaching_rms": 0.0032383314577264692, + "traffic_rms": 0.07053831711802178 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.6427491184886094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5468975658595973e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01179781578136396, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16286016123771668 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03283748775720596, + "mean_feedback_forward_cosine": 0.9997327346955577, + "mean_feedback_forward_relative_error": 0.022705907362603372, + "min_feedback_forward_cosine": 0.9994620680809021 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031856518329577478, + "instruction_rms": 0.0031856518327997496, + "raw_apical_rms": 0.07086310605920623, + "teaching_rms": 0.0031856518329577478, + "traffic_rms": 0.07078882961010506 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.042470974103708e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5447916899787632e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013604974040193276, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16662931219100952 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03161013498902321, + "mean_feedback_forward_cosine": 0.9997522350280516, + "mean_feedback_forward_relative_error": 0.021860082483580037, + "min_feedback_forward_cosine": 0.9995020627975464 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00315390776011399, + "instruction_rms": 0.0031539077604541403, + "raw_apical_rms": 0.07085140910107751, + "teaching_rms": 0.00315390776011399, + "traffic_rms": 0.07077858091558684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.616428839199216e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542393493814238e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012310409672314714, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1531004501914978 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0315987691283226, + "mean_feedback_forward_cosine": 0.999752402305603, + "mean_feedback_forward_relative_error": 0.02185060578067937, + "min_feedback_forward_cosine": 0.999502420425415 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0023233509040279073, + "instruction_rms": 0.0023233509036406463, + "raw_apical_rms": 0.0707654348805652, + "teaching_rms": 0.0023233509040279073, + "traffic_rms": 0.07072494607504694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.824864336365863e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5250846180992398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011755987782357761, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.087803178024292 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031592871993780136, + "mean_feedback_forward_cosine": 0.9997525099785097, + "mean_feedback_forward_relative_error": 0.021844573679470247, + "min_feedback_forward_cosine": 0.9995025396347046 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002086301752220019, + "instruction_rms": 0.002086301752638763, + "raw_apical_rms": 0.07076359193145179, + "teaching_rms": 0.002086301752220019, + "traffic_rms": 0.07073055681585723 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3719278863818545e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5290625533644456e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011577810160412101, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06659797365188598 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0315878689289093, + "mean_feedback_forward_cosine": 0.9997526099604945, + "mean_feedback_forward_relative_error": 0.021839028044093038, + "min_feedback_forward_cosine": 0.9995027780532837 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001918077645467979, + "instruction_rms": 0.0019180776456162404, + "raw_apical_rms": 0.07080159516314757, + "teaching_rms": 0.001918077645467979, + "traffic_rms": 0.07077324073624926 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.739856507285367e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528011983324446e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011458038375340998, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.055139399154186246 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03158288821578026, + "mean_feedback_forward_cosine": 0.9997526868697135, + "mean_feedback_forward_relative_error": 0.02183374678415637, + "min_feedback_forward_cosine": 0.9995028972625732 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019071023736372602, + "instruction_rms": 0.001907102373116384, + "raw_apical_rms": 0.07077306510584018, + "teaching_rms": 0.0019071023736372602, + "traffic_rms": 0.07074480243596765 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.7397786201154304e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5252243042118834e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01130050615419107, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.050355927820801734 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031578391790390015, + "mean_feedback_forward_cosine": 0.9997527733925851, + "mean_feedback_forward_relative_error": 0.021828834177745927, + "min_feedback_forward_cosine": 0.999502956867218 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018590293028376338, + "instruction_rms": 0.00185902930266732, + "raw_apical_rms": 0.07078169367004526, + "teaching_rms": 0.0018590293028376338, + "traffic_rms": 0.07075462945196666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.4625362960459825e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5285456262488057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011481049939528766, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04517262459874153 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031573761254549026, + "mean_feedback_forward_cosine": 0.999752854147265, + "mean_feedback_forward_relative_error": 0.021824016896707397, + "min_feedback_forward_cosine": 0.9995031356811523 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018453101634222267, + "instruction_rms": 0.0018453101629226716, + "raw_apical_rms": 0.07078147952202035, + "teaching_rms": 0.0018453101634222267, + "traffic_rms": 0.07075500902065815 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.7035672306728884e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5216579613205712e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011481601696436076, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.041840342236757276 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03156920149922371, + "mean_feedback_forward_cosine": 0.9997529406701365, + "mean_feedback_forward_relative_error": 0.02181929447537949, + "min_feedback_forward_cosine": 0.9995032548904419 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017880301606806024, + "instruction_rms": 0.001788030160392268, + "raw_apical_rms": 0.07080753844098246, + "teaching_rms": 0.0017880301606806024, + "traffic_rms": 0.07078193147695483 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.7850171824848076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5179552499449808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011930645049259392, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.038340160447359085 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03156426548957825, + "mean_feedback_forward_cosine": 0.9997530175793555, + "mean_feedback_forward_relative_error": 0.021814521942888538, + "min_feedback_forward_cosine": 0.999503493309021 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018164985512064019, + "instruction_rms": 0.0018164985514785225, + "raw_apical_rms": 0.07081565415120102, + "teaching_rms": 0.0018164985512064019, + "traffic_rms": 0.0707890240884415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.641485062961692e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.523815231951257e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01158515836348895, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.0374163833296299 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031559910625219345, + "mean_feedback_forward_cosine": 0.9997531060249575, + "mean_feedback_forward_relative_error": 0.021810060547244166, + "min_feedback_forward_cosine": 0.9995035529136658 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017306366105848245, + "instruction_rms": 0.0017306366103148378, + "raw_apical_rms": 0.07083518009522981, + "teaching_rms": 0.0017306366105848245, + "traffic_rms": 0.07081101047000242 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.547128267835433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5210304677114488e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011715913631631722, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.03402915516614914 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03155505284667015, + "mean_feedback_forward_cosine": 0.9997531790887156, + "mean_feedback_forward_relative_error": 0.021805410769077077, + "min_feedback_forward_cosine": 0.9995036125183105 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018028494172434846, + "instruction_rms": 0.0018028494170743582, + "raw_apical_rms": 0.07082326782268145, + "teaching_rms": 0.0018028494172434846, + "traffic_rms": 0.07079719189833561 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.701535954472092e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5253110590846376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0116941067721118, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.033918422058522704 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03155054897069931, + "mean_feedback_forward_cosine": 0.9997532502297433, + "mean_feedback_forward_relative_error": 0.021800997636971936, + "min_feedback_forward_cosine": 0.9995037317276001 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001695522417788297, + "instruction_rms": 0.0016955224176175118, + "raw_apical_rms": 0.07081304541149969, + "teaching_rms": 0.001695522417788297, + "traffic_rms": 0.07078941366704658 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.8828908916220826e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5330936330274074e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011615912357428653, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.0304179566141963 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031545937061309814, + "mean_feedback_forward_cosine": 0.9997533405980756, + "mean_feedback_forward_relative_error": 0.02179660527936874, + "min_feedback_forward_cosine": 0.9995039105415344 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016762094801257802, + "instruction_rms": 0.0016762094801564974, + "raw_apical_rms": 0.07078776126405462, + "teaching_rms": 0.0016762094801257802, + "traffic_rms": 0.07076422476020912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.814752330162264e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537455768209822e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011348902066832332, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.028853442577719688 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03154109790921211, + "mean_feedback_forward_cosine": 0.9997534155845642, + "mean_feedback_forward_relative_error": 0.021792131761509563, + "min_feedback_forward_cosine": 0.9995039701461792 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017517046770167588, + "instruction_rms": 0.0017517046768157352, + "raw_apical_rms": 0.07071229405494915, + "teaching_rms": 0.0017517046770167588, + "traffic_rms": 0.07068692920553031 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.9297853738371487e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.525912084264886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011382223286464578, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.029125761003494264 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03153675049543381, + "mean_feedback_forward_cosine": 0.9997534809574005, + "mean_feedback_forward_relative_error": 0.021787961389148427, + "min_feedback_forward_cosine": 0.9995041489601135 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015876705038328752, + "instruction_rms": 0.001587670503565514, + "raw_apical_rms": 0.07064813430811864, + "teaching_rms": 0.0015876705038328752, + "traffic_rms": 0.07062696624907579 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.896137457104487e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.525642220546241e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011553039736056642, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.025205514112114906 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03153231367468834, + "mean_feedback_forward_cosine": 0.9997535809393852, + "mean_feedback_forward_relative_error": 0.021783713942333575, + "min_feedback_forward_cosine": 0.9995042681694031 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016591918330534736, + "instruction_rms": 0.001659191833391364, + "raw_apical_rms": 0.07066638729512363, + "teaching_rms": 0.0016591918330534736, + "traffic_rms": 0.0706434317968168 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.081116600218593e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522773624800793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011410539204891235, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.026160163542032242 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03152792900800705, + "mean_feedback_forward_cosine": 0.9997536520804128, + "mean_feedback_forward_relative_error": 0.021779457062122323, + "min_feedback_forward_cosine": 0.9995045065879822 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001635147070197642, + "instruction_rms": 0.0016351470699119863, + "raw_apical_rms": 0.07063231208621541, + "teaching_rms": 0.001635147070197642, + "traffic_rms": 0.07060969015503646 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.947635107692804e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5290371258718755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011698994184058217, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.02423469348013401 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03152330592274666, + "mean_feedback_forward_cosine": 0.9997537136077881, + "mean_feedback_forward_relative_error": 0.021775202344982855, + "min_feedback_forward_cosine": 0.999504566192627 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016578184965252074, + "instruction_rms": 0.0016578184965053524, + "raw_apical_rms": 0.07062680307400425, + "teaching_rms": 0.0016578184965252074, + "traffic_rms": 0.07060324110481668 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.4363422208325574e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529187417623237e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011549385945445856, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.023876547414064406 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03151863440871239, + "mean_feedback_forward_cosine": 0.9997538135897729, + "mean_feedback_forward_relative_error": 0.02177095501833866, + "min_feedback_forward_cosine": 0.999504566192627 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015907520566565076, + "instruction_rms": 0.0015907520569681812, + "raw_apical_rms": 0.07062726369724308, + "teaching_rms": 0.0015907520566565076, + "traffic_rms": 0.0706050703322031 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.327001335013847e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5275007657435867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011653882141844858, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.02222636599600315 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031513750553131104, + "mean_feedback_forward_cosine": 0.9997538885762615, + "mean_feedback_forward_relative_error": 0.021766589323599494, + "min_feedback_forward_cosine": 0.9995046854019165 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016582706109966375, + "instruction_rms": 0.0016582706109674029, + "raw_apical_rms": 0.07064716101403573, + "teaching_rms": 0.0016582706109966375, + "traffic_rms": 0.07062343763721211 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.2123382115732966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5217520753111575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011575557945469163, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.02295726150751114 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031509749591350555, + "mean_feedback_forward_cosine": 0.99975396356275, + "mean_feedback_forward_relative_error": 0.021762791345076214, + "min_feedback_forward_cosine": 0.9995049238204956 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014430583172411718, + "instruction_rms": 0.0014430583176640059, + "raw_apical_rms": 0.07064072797242527, + "teaching_rms": 0.0014430583172411718, + "traffic_rms": 0.0706224000119091 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.149996257751809e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5247631730813373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01190286897094364, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.018640852463841437 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031505174934864044, + "mean_feedback_forward_cosine": 0.9997540000946291, + "mean_feedback_forward_relative_error": 0.02175876636418604, + "min_feedback_forward_cosine": 0.9995050430297852 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015259132571148375, + "instruction_rms": 0.0015259132572187795, + "raw_apical_rms": 0.07062594995389963, + "teaching_rms": 0.0015259132571148375, + "traffic_rms": 0.07060563620985064 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 5.964832894278516e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5257133712340473e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011830879638842778, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.019138810096085073 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03150048851966858, + "mean_feedback_forward_cosine": 0.9997541019993443, + "mean_feedback_forward_relative_error": 0.021754628092411063, + "min_feedback_forward_cosine": 0.9995050430297852 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015848297433587592, + "instruction_rms": 0.0015848297430128633, + "raw_apical_rms": 0.07062075151293705, + "teaching_rms": 0.0015848297433587592, + "traffic_rms": 0.07059855406855263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.609144355072203e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5294070388716787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011680275251620733, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01945107658326626 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031495921313762665, + "mean_feedback_forward_cosine": 0.9997541616039891, + "mean_feedback_forward_relative_error": 0.02175049788709129, + "min_feedback_forward_cosine": 0.9995052218437195 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014838732998784865, + "instruction_rms": 0.0014838732998472648, + "raw_apical_rms": 0.07063766199187763, + "teaching_rms": 0.0014838732998784865, + "traffic_rms": 0.07061766313094757 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.405471543122985e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5277802982346156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011511931936851679, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.018230360715687274 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03149140998721123, + "mean_feedback_forward_cosine": 0.9997542635087044, + "mean_feedback_forward_relative_error": 0.02174653616102953, + "min_feedback_forward_cosine": 0.9995054006576538 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001483081299368288, + "instruction_rms": 0.001483081299277758, + "raw_apical_rms": 0.07055642986565698, + "teaching_rms": 0.001483081299368288, + "traffic_rms": 0.07053611410287747 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.846123502488808e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53047037930489e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01160453884440212, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.01801989272594452 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03148650750517845, + "mean_feedback_forward_cosine": 0.9997543269588102, + "mean_feedback_forward_relative_error": 0.021742422268876144, + "min_feedback_forward_cosine": 0.9995054006576538 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014884369870363553, + "instruction_rms": 0.001488436987233521, + "raw_apical_rms": 0.07054572070927104, + "teaching_rms": 0.0014884369870363553, + "traffic_rms": 0.07052549174688853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.718665636005426e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5318834129771767e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011617521675298665, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01730287718936801 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03148270398378372, + "mean_feedback_forward_cosine": 0.9997543923316463, + "mean_feedback_forward_relative_error": 0.021738420090367718, + "min_feedback_forward_cosine": 0.9995055198669434 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014378686815881324, + "instruction_rms": 0.001437868681970199, + "raw_apical_rms": 0.07055516889107566, + "teaching_rms": 0.0014378686815881324, + "traffic_rms": 0.07053606851764507 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.979782275427169e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528494848104146e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011662366378582107, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01609927766799927 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03147909417748451, + "mean_feedback_forward_cosine": 0.9997544577044826, + "mean_feedback_forward_relative_error": 0.021734440549006386, + "min_feedback_forward_cosine": 0.9995056390762329 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014434023992191641, + "instruction_rms": 0.0014434023986252685, + "raw_apical_rms": 0.07055555375402328, + "teaching_rms": 0.0014434023992191641, + "traffic_rms": 0.07053664169991297 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.7707764849747036e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5267121645867264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011707232682734508, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.016113468016982078 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03147529810667038, + "mean_feedback_forward_cosine": 0.9997545615319283, + "mean_feedback_forward_relative_error": 0.021730275884751352, + "min_feedback_forward_cosine": 0.9995058178901672 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001455725419073, + "instruction_rms": 0.0014557254187307872, + "raw_apical_rms": 0.07051318189153234, + "teaching_rms": 0.001455725419073, + "traffic_rms": 0.07049314415652437 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.442953382977066e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.523862240188352e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01178436885542076, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.016095425399839876 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031471483409404755, + "mean_feedback_forward_cosine": 0.9997546442093388, + "mean_feedback_forward_relative_error": 0.02172613839408563, + "min_feedback_forward_cosine": 0.9995059370994568 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001468184035216762, + "instruction_rms": 0.0014681840349298986, + "raw_apical_rms": 0.07048467015219818, + "teaching_rms": 0.001468184035216762, + "traffic_rms": 0.07046493453509667 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.5252673547556697e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.52245917175342e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011577974801195734, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.01604898440271616 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031467828899621964, + "mean_feedback_forward_cosine": 0.9997547384231321, + "mean_feedback_forward_relative_error": 0.021722148593154647, + "min_feedback_forward_cosine": 0.9995059967041016 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013914267274000695, + "instruction_rms": 0.0013914267275130041, + "raw_apical_rms": 0.07045413473438684, + "teaching_rms": 0.0013914267274000695, + "traffic_rms": 0.070435154277 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.960167387935144e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5250880902871854e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01161515458950019, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.01481641760468483 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03146391361951828, + "mean_feedback_forward_cosine": 0.9997548095641597, + "mean_feedback_forward_relative_error": 0.021717941914234432, + "min_feedback_forward_cosine": 0.9995062351226807 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014164737671420239, + "instruction_rms": 0.0014164737672724631, + "raw_apical_rms": 0.07045084255377658, + "teaching_rms": 0.0014164737671420239, + "traffic_rms": 0.07043102485962732 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.192543912040492e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5248540174451093e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011450585284064223, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.014877579310536385 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031460799276828766, + "mean_feedback_forward_cosine": 0.9997548537869607, + "mean_feedback_forward_relative_error": 0.021714498364036123, + "min_feedback_forward_cosine": 0.9995061755180359 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012618732316198448, + "instruction_rms": 0.0012618732319364578, + "raw_apical_rms": 0.07043261242790871, + "teaching_rms": 0.0012618732316198448, + "traffic_rms": 0.07041709060142269 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 6.379265471423423e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5249243004196957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011566134962821875, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.012776756524145604 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03145720809698105, + "mean_feedback_forward_cosine": 0.9997549499234846, + "mean_feedback_forward_relative_error": 0.02171059481559261, + "min_feedback_forward_cosine": 0.9995062351226807 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013324146890190173, + "instruction_rms": 0.0013324146890174922, + "raw_apical_rms": 0.07042891036111297, + "teaching_rms": 0.0013324146890190173, + "traffic_rms": 0.07041160000418968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.440669698624333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5300426206559954e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011742498019665766, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013463607396483421 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03145337849855423, + "mean_feedback_forward_cosine": 0.999755036446356, + "mean_feedback_forward_relative_error": 0.02170649945976273, + "min_feedback_forward_cosine": 0.9995064735412598 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014016867772730533, + "instruction_rms": 0.0014016867769787247, + "raw_apical_rms": 0.07037403322697358, + "teaching_rms": 0.0014016867772730533, + "traffic_rms": 0.07035478451789434 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 5.130643998541018e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5222852045398864e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011743140404652274, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.01355147111326456 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031449656933546066, + "mean_feedback_forward_cosine": 0.9997550979737313, + "mean_feedback_forward_relative_error": 0.021702520834702636, + "min_feedback_forward_cosine": 0.9995065927505493 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013293457666306403, + "instruction_rms": 0.001329345766659357, + "raw_apical_rms": 0.0703187473802673, + "teaching_rms": 0.0013293457666306403, + "traffic_rms": 0.07030111325374616 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.708783324896103e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527145619713989e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011787931924955846, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.01294834453187883 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031446296721696854, + "mean_feedback_forward_cosine": 0.9997551902647941, + "mean_feedback_forward_relative_error": 0.021698803205283418, + "min_feedback_forward_cosine": 0.9995067119598389 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001273240062966235, + "instruction_rms": 0.0012732400631295484, + "raw_apical_rms": 0.07031436334846808, + "teaching_rms": 0.001273240062966235, + "traffic_rms": 0.07029834416585731 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.1766870850779014e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5252506512422616e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011681072992791103, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.01216580724567175 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031442634761333466, + "mean_feedback_forward_cosine": 0.9997552479467084, + "mean_feedback_forward_relative_error": 0.021694860391078458, + "min_feedback_forward_cosine": 0.9995068311691284 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013001379488674402, + "instruction_rms": 0.001300137948847344, + "raw_apical_rms": 0.0703550425587763, + "teaching_rms": 0.0013001379488674402, + "traffic_rms": 0.07033806633157637 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.291127536111162e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.520106225759273e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011688400551326618, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.012553872425705194 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031438663601875305, + "mean_feedback_forward_cosine": 0.9997553421605018, + "mean_feedback_forward_relative_error": 0.02169070385336395, + "min_feedback_forward_cosine": 0.9995068311691284 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013336208403755298, + "instruction_rms": 0.0013336208402559818, + "raw_apical_rms": 0.07036283304687435, + "teaching_rms": 0.0013336208403755298, + "traffic_rms": 0.0703448289272187 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.578428658419398e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53044558532039e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011879580117056813, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012370717002600432 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03143467381596565, + "mean_feedback_forward_cosine": 0.9997554209924513, + "mean_feedback_forward_relative_error": 0.021686465554540197, + "min_feedback_forward_cosine": 0.9995070695877075 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013706312123222073, + "instruction_rms": 0.0013706312123766054, + "raw_apical_rms": 0.07035709887665909, + "teaching_rms": 0.0013706312123222073, + "traffic_rms": 0.07033896256840985 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.476185466728566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5246471801487927e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011824206816571706, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012631570869535208 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031431134790182114, + "mean_feedback_forward_cosine": 0.9997554902107485, + "mean_feedback_forward_relative_error": 0.021682552001889673, + "min_feedback_forward_cosine": 0.9995070695877075 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012590432912691625, + "instruction_rms": 0.0012590432913530197, + "raw_apical_rms": 0.07031875204929189, + "teaching_rms": 0.0012590432912691625, + "traffic_rms": 0.07030294213585257 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.9685294101209365e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5225175528654364e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011784858272875514, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.01128523970939219 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031426865607500076, + "mean_feedback_forward_cosine": 0.9997556113427685, + "mean_feedback_forward_relative_error": 0.02167809260408244, + "min_feedback_forward_cosine": 0.9995072484016418 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001359515134284952, + "instruction_rms": 0.0013595151341748465, + "raw_apical_rms": 0.070283139455583, + "teaching_rms": 0.001359515134284952, + "traffic_rms": 0.07026412445008359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3798621618407196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5317886108972528e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012017519705804336, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.012544265588521957 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03142315894365311, + "mean_feedback_forward_cosine": 0.9997556632564913, + "mean_feedback_forward_relative_error": 0.021674015394021428, + "min_feedback_forward_cosine": 0.9995073676109314 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012819728074822169, + "instruction_rms": 0.0012819728075502469, + "raw_apical_rms": 0.07027961951572126, + "teaching_rms": 0.0012819728074822169, + "traffic_rms": 0.07026295579769154 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3103984381787086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5328682217318875e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011752616225298043, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.011569785919934511 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03141937404870987, + "mean_feedback_forward_cosine": 0.9997557555475542, + "mean_feedback_forward_relative_error": 0.021669918521037026, + "min_feedback_forward_cosine": 0.9995075464248657 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012574740021123855, + "instruction_rms": 0.0012574740024526203, + "raw_apical_rms": 0.07028106837953021, + "teaching_rms": 0.0012574740021123855, + "traffic_rms": 0.07026506957979557 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.7281619569284885e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359786923545134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011781703164463598, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011055297872126103 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03141534700989723, + "mean_feedback_forward_cosine": 0.9997558266885819, + "mean_feedback_forward_relative_error": 0.02166577610337446, + "min_feedback_forward_cosine": 0.999507486820221 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001269485828808143, + "instruction_rms": 0.0012694858287749988, + "raw_apical_rms": 0.07028588119028038, + "teaching_rms": 0.001269485828808143, + "traffic_rms": 0.07026939352785953 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.5105004176093644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5281909345459418e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01201613598204956, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.011340039749145507 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03141167759895325, + "mean_feedback_forward_cosine": 0.9997559382069495, + "mean_feedback_forward_relative_error": 0.021661755268371875, + "min_feedback_forward_cosine": 0.9995077252388 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012045931843313916, + "instruction_rms": 0.001204593184421599, + "raw_apical_rms": 0.07029184532932393, + "teaching_rms": 0.0012045931843313916, + "traffic_rms": 0.07027640189253148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 5.3829573687380616e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522316529583776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011849451581964304, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010319001391977072 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03140778839588165, + "mean_feedback_forward_cosine": 0.9997559901206724, + "mean_feedback_forward_relative_error": 0.02165760123921979, + "min_feedback_forward_cosine": 0.9995076656341553 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012605488668036228, + "instruction_rms": 0.001260548867102661, + "raw_apical_rms": 0.0703019797435531, + "teaching_rms": 0.0012605488668036228, + "traffic_rms": 0.07028575949204856 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.307186785013073e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.52301780433182e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011616746506788761, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010907049433887005 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031404104083776474, + "mean_feedback_forward_cosine": 0.9997560516480477, + "mean_feedback_forward_relative_error": 0.021653504516448704, + "min_feedback_forward_cosine": 0.9995078444480896 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012157758465225696, + "instruction_rms": 0.0012157758466930155, + "raw_apical_rms": 0.07026565535405105, + "teaching_rms": 0.0012157758465225696, + "traffic_rms": 0.07025011098288815 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.8966334986744187e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5285786976708692e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01191025353698056, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010033720988929271 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139985725283623, + "mean_feedback_forward_cosine": 0.9997561612436848, + "mean_feedback_forward_relative_error": 0.021649181362121336, + "min_feedback_forward_cosine": 0.9995079636573792 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001216893373104312, + "instruction_rms": 0.0012168933727593268, + "raw_apical_rms": 0.07026047437396069, + "teaching_rms": 0.001216893373104312, + "traffic_rms": 0.07024403139859071 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.6607688375388534e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536955170363654e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011770401000829997, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.010272326285541058 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031395841389894485, + "mean_feedback_forward_cosine": 0.999756234307443, + "mean_feedback_forward_relative_error": 0.02164496568542334, + "min_feedback_forward_cosine": 0.9995081424713135 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012171149515118225, + "instruction_rms": 0.0012171149513816762, + "raw_apical_rms": 0.07023720382268496, + "teaching_rms": 0.0012171149515118225, + "traffic_rms": 0.07022166614737066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.481043885424697e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.531877272456318e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011703893236434766, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.010315310284942389 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313919372856617, + "mean_feedback_forward_cosine": 0.9997563342894277, + "mean_feedback_forward_relative_error": 0.02164069424954153, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012118312013621908, + "instruction_rms": 0.001211831201467192, + "raw_apical_rms": 0.07024763818861808, + "teaching_rms": 0.0012118312013621908, + "traffic_rms": 0.0702320355387216 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.94944289389332e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5258519139627042e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011885435114700417, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009989295650720596 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139188885688782, + "mean_feedback_forward_cosine": 0.9997563265985058, + "mean_feedback_forward_relative_error": 0.02164059446283406, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001132558731999788, + "instruction_rms": 0.0011325587324076524, + "raw_apical_rms": 0.0702130866659378, + "teaching_rms": 0.001132558731999788, + "traffic_rms": 0.0701986342614779 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.750136912117341e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5251534036711087e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011795561430422253, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.008952871725261212 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139185905456543, + "mean_feedback_forward_cosine": 0.9997563650531154, + "mean_feedback_forward_relative_error": 0.02164050899145584, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010287751242244847, + "instruction_rms": 0.0010287751239929643, + "raw_apical_rms": 0.07021523367502283, + "teaching_rms": 0.0010287751242244847, + "traffic_rms": 0.07020311277239633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.014024395677552e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528264417870819e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011757748559011854, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008012445749863982 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139183297753334, + "mean_feedback_forward_cosine": 0.9997563400576192, + "mean_feedback_forward_relative_error": 0.021640425157402794, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009815886566740938, + "instruction_rms": 0.00098158865682698, + "raw_apical_rms": 0.07021620195178999, + "teaching_rms": 0.0009815886566740938, + "traffic_rms": 0.07020548496672611 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.9341153385285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5315856392138565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011709095120630406, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.007590246652066707 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139180690050125, + "mean_feedback_forward_cosine": 0.9997563362121582, + "mean_feedback_forward_relative_error": 0.021640341984288346, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001037752511971386, + "instruction_rms": 0.0010377525120009354, + "raw_apical_rms": 0.07021965644850046, + "teaching_rms": 0.001037752511971386, + "traffic_rms": 0.07020762901755545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.1557854046914144e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.52778257126182e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011671761692863963, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.008242070003896951 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391777098178864, + "mean_feedback_forward_cosine": 0.9997563381348887, + "mean_feedback_forward_relative_error": 0.021640257143806064, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010358130376364439, + "instruction_rms": 0.0010358130375094992, + "raw_apical_rms": 0.07021829007518812, + "teaching_rms": 0.0010358130376364439, + "traffic_rms": 0.0702060549591956 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.7380129214170665e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5277152994051676e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011867800848287426, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.00832056779332459 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391751021146774, + "mean_feedback_forward_cosine": 0.9997563458258106, + "mean_feedback_forward_relative_error": 0.02164017695993666, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000912207601210762, + "instruction_rms": 0.0009122076011691182, + "raw_apical_rms": 0.07021351585483629, + "teaching_rms": 0.000912207601210762, + "traffic_rms": 0.07020382076460356 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.746044351440105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5273115013650083e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011687577159252917, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.006752948218658567 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139171749353409, + "mean_feedback_forward_cosine": 0.9997563400576192, + "mean_feedback_forward_relative_error": 0.02164009736189919, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000996774933101697, + "instruction_rms": 0.0009967749328507287, + "raw_apical_rms": 0.07021260599151836, + "teaching_rms": 0.000996774933101697, + "traffic_rms": 0.07020106486913297 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.84565738645415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5303641551369146e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011828175358131987, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.0076510861584544185 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391698867082596, + "mean_feedback_forward_cosine": 0.9997563323666973, + "mean_feedback_forward_relative_error": 0.021640019416208228, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009428604233684526, + "instruction_rms": 0.0009428604234087362, + "raw_apical_rms": 0.07021038914772855, + "teaching_rms": 0.0009428604233684526, + "traffic_rms": 0.07020014992619868 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.8097664517854355e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5303295507089353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011679169371461913, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.00707207479968667 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139167279005051, + "mean_feedback_forward_cosine": 0.9997563496712716, + "mean_feedback_forward_relative_error": 0.02163994198124255, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009265202989265416, + "instruction_rms": 0.0009265202985500273, + "raw_apical_rms": 0.07020828628755175, + "teaching_rms": 0.0009265202989265416, + "traffic_rms": 0.07019812365940475 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.8420744747090794e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5303389260400146e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011613886481925927, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007113533585369587 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139164671301842, + "mean_feedback_forward_cosine": 0.9997563573621935, + "mean_feedback_forward_relative_error": 0.021639861196519866, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010189411475243957, + "instruction_rms": 0.0010189411477495233, + "raw_apical_rms": 0.07020948590509299, + "teaching_rms": 0.0010189411475243957, + "traffic_rms": 0.07019731298623119 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.564643774074284e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5289907285707224e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011705157023293335, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007799769958555698 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139162436127663, + "mean_feedback_forward_cosine": 0.9997563304439667, + "mean_feedback_forward_relative_error": 0.021639785954668637, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008879372602193249, + "instruction_rms": 0.0008879372602216135, + "raw_apical_rms": 0.07021130317749433, + "teaching_rms": 0.0008879372602193249, + "traffic_rms": 0.07020222280655782 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.843028378331837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5294956589401724e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011860261194473797, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.006647673753872514 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139159828424454, + "mean_feedback_forward_cosine": 0.9997563419803497, + "mean_feedback_forward_relative_error": 0.021639709811537497, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009080461846485905, + "instruction_rms": 0.0009080461846861004, + "raw_apical_rms": 0.0702107428884489, + "teaching_rms": 0.0009080461846485905, + "traffic_rms": 0.07020125201700089 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.5260819920022186e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5286520750964776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011713258880136204, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.006807348629161715 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139157220721245, + "mean_feedback_forward_cosine": 0.9997563439030801, + "mean_feedback_forward_relative_error": 0.021639633232787732, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009873041037586194, + "instruction_rms": 0.000987304104052906, + "raw_apical_rms": 0.07021038552535869, + "teaching_rms": 0.0009873041037586194, + "traffic_rms": 0.07019870663016894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.4286348810373966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5273341589930466e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01181477957551801, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007489908771365881 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139154613018036, + "mean_feedback_forward_cosine": 0.999756351594002, + "mean_feedback_forward_relative_error": 0.02163955593301404, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009085053031699389, + "instruction_rms": 0.0009085053032427112, + "raw_apical_rms": 0.07020804708510761, + "teaching_rms": 0.0009085053031699389, + "traffic_rms": 0.07019822440858917 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.6827988597897274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5302806209249857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011752074580298671, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007063958764821291 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139152005314827, + "mean_feedback_forward_cosine": 0.9997563439030801, + "mean_feedback_forward_relative_error": 0.02163947686072319, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008957087453228049, + "instruction_rms": 0.0008957087453178347, + "raw_apical_rms": 0.07020699042292007, + "teaching_rms": 0.0008957087453228049, + "traffic_rms": 0.07019785084546254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.128126818159217e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528924564910661e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011658755907931568, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.006880758673548698 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139149397611618, + "mean_feedback_forward_cosine": 0.9997563592849239, + "mean_feedback_forward_relative_error": 0.021639399170394865, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009090331567331726, + "instruction_rms": 0.0009090331568032205, + "raw_apical_rms": 0.07020432885938041, + "teaching_rms": 0.0009090331567331726, + "traffic_rms": 0.07019504583687272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.5084355804192455e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5276196264429482e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012045960536030787, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.006840768229588866 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139146789908409, + "mean_feedback_forward_cosine": 0.9997563573621935, + "mean_feedback_forward_relative_error": 0.021639324319098267, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009608543221332888, + "instruction_rms": 0.0009608543223451664, + "raw_apical_rms": 0.07020496352440089, + "teaching_rms": 0.0009608543221332888, + "traffic_rms": 0.07019431420061167 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.8637580246358993e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5305563103426655e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01174591127526034, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007100784926041961 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391438096761703, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.021639249813292297, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008600077225515398, + "instruction_rms": 0.0008600077226331852, + "raw_apical_rms": 0.07020018222045596, + "teaching_rms": 0.0008600077225515398, + "traffic_rms": 0.07019151401309603 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.2788581566810535e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5313721966525293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011715586165079875, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006382754127047956 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139141574501991, + "mean_feedback_forward_cosine": 0.999756351594002, + "mean_feedback_forward_relative_error": 0.02163917297917989, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009174612628861559, + "instruction_rms": 0.0009174612626906001, + "raw_apical_rms": 0.07020379219663715, + "teaching_rms": 0.0009174612628861559, + "traffic_rms": 0.07019412865723207 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.710185936256494e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.526453616736127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011687319644047654, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006855643994361162 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391389667987823, + "mean_feedback_forward_cosine": 0.9997563496712716, + "mean_feedback_forward_relative_error": 0.02163909447769965, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008852335151270924, + "instruction_rms": 0.0008852335147614269, + "raw_apical_rms": 0.07019929138721527, + "teaching_rms": 0.0008852335151270924, + "traffic_rms": 0.07018995991445826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.6087851757192766e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.526767767201796e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011747481731209234, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.006547002290561796 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139136731624603, + "mean_feedback_forward_cosine": 0.9997563708213068, + "mean_feedback_forward_relative_error": 0.02163902004700034, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009002472914467452, + "instruction_rms": 0.0009002472915304042, + "raw_apical_rms": 0.07019637018038145, + "teaching_rms": 0.0009002472914467452, + "traffic_rms": 0.07018689896652486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.400720499120325e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5280779675430417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011722185639443923, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0065838918869942425 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391337513923645, + "mean_feedback_forward_cosine": 0.9997563419803497, + "mean_feedback_forward_relative_error": 0.02163894333305859, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008443229673745806, + "instruction_rms": 0.0008443229674634042, + "raw_apical_rms": 0.0701936966019172, + "teaching_rms": 0.0008443229673745806, + "traffic_rms": 0.07018540330225034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.631268086223827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5279407298314868e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011712241480544405, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006101330086141825 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391315162181854, + "mean_feedback_forward_cosine": 0.9997563862031505, + "mean_feedback_forward_relative_error": 0.021638870089044495, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00087445066005554, + "instruction_rms": 0.000874450660164008, + "raw_apical_rms": 0.07019014315623327, + "teaching_rms": 0.00087445066005554, + "traffic_rms": 0.07018040313402471 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.5814385174480776e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5275774595678575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011892064224213659, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0065509532841295 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139128535985947, + "mean_feedback_forward_cosine": 0.999756347748541, + "mean_feedback_forward_relative_error": 0.021638795898686492, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000849687844803133, + "instruction_rms": 0.0008496878445148154, + "raw_apical_rms": 0.07018789212212574, + "teaching_rms": 0.000849687844803133, + "traffic_rms": 0.0701788102321727 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3653173520397104e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527847863993578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011984499768589538, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006231142134666443 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139127045869827, + "mean_feedback_forward_cosine": 0.9997563612076544, + "mean_feedback_forward_relative_error": 0.021638725223320145, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008935700128148828, + "instruction_rms": 0.000893570012921475, + "raw_apical_rms": 0.0701855992409011, + "teaching_rms": 0.0008935700128148828, + "traffic_rms": 0.07017611766983055 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.833876505154876e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5234644800587153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011750247721697636, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0066043989513069395 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139125183224678, + "mean_feedback_forward_cosine": 0.9997563804349592, + "mean_feedback_forward_relative_error": 0.021638656230342966, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008790336939580951, + "instruction_rms": 0.0008790336938571931, + "raw_apical_rms": 0.07018557377674385, + "teaching_rms": 0.0008790336939580951, + "traffic_rms": 0.07017608095416995 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.834886186382391e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5316402739340936e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011966530677576194, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006521816009432077 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139122575521469, + "mean_feedback_forward_cosine": 0.9997563592849239, + "mean_feedback_forward_relative_error": 0.021638581063598394, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008805740563806255, + "instruction_rms": 0.0008805740562376627, + "raw_apical_rms": 0.070187052173234, + "teaching_rms": 0.0008805740563806255, + "traffic_rms": 0.07017778428289262 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.317117057757059e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528882379642278e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011875181198551011, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006496892674937844 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031391192227602005, + "mean_feedback_forward_cosine": 0.9997563727440373, + "mean_feedback_forward_relative_error": 0.02163850124024095, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009164387661969439, + "instruction_rms": 0.000916438766474656, + "raw_apical_rms": 0.07018409380399021, + "teaching_rms": 0.0009164387661969439, + "traffic_rms": 0.07017418972022275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.570502281670997e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.526979933350636e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011674213681022505, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0063994974429905415 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139117360115051, + "mean_feedback_forward_cosine": 0.9997563650531154, + "mean_feedback_forward_relative_error": 0.02163842710996828, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008503840226223957, + "instruction_rms": 0.0008503840226210507, + "raw_apical_rms": 0.07018257020935656, + "teaching_rms": 0.0008503840226223957, + "traffic_rms": 0.07017403118147252 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.886266952690676e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527368989092424e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011901159616179633, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006223728241473436 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139114752411842, + "mean_feedback_forward_cosine": 0.9997563631303849, + "mean_feedback_forward_relative_error": 0.021638350696453163, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008786705914125536, + "instruction_rms": 0.0008786705917428955, + "raw_apical_rms": 0.07018214116704091, + "teaching_rms": 0.0008786705914125536, + "traffic_rms": 0.0701728942376601 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.198854597348145e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528797597143655e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011943473090484474, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006245601629540324 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139113262295723, + "mean_feedback_forward_cosine": 0.9997563535167325, + "mean_feedback_forward_relative_error": 0.021638275454601934, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008749180366986777, + "instruction_rms": 0.0008749180366827683, + "raw_apical_rms": 0.07018262203755016, + "teaching_rms": 0.0008749180366986777, + "traffic_rms": 0.07017305205320808 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.398657083015678e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527683617348162e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011861382040144832, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006101286468356848 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139111399650574, + "mean_feedback_forward_cosine": 0.9997563688985763, + "mean_feedback_forward_relative_error": 0.02163820138441459, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008315546744591696, + "instruction_rms": 0.0008315546744404981, + "raw_apical_rms": 0.07018149744220303, + "teaching_rms": 0.0008315546744591696, + "traffic_rms": 0.0701731126295513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.169811518295319e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5302107165465284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011950335507668353, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006009815453365445 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139108419418335, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.021638124790643493, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008876319843738879, + "instruction_rms": 0.0008876319845756654, + "raw_apical_rms": 0.07018309484011162, + "teaching_rms": 0.0008876319843738879, + "traffic_rms": 0.0701731981207293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.687281401378641e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.528633583326007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01212179218307607, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0064534566259384155 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139106556773186, + "mean_feedback_forward_cosine": 0.9997563785122286, + "mean_feedback_forward_relative_error": 0.02163805108096811, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008352924901350643, + "instruction_rms": 0.0008352924900780053, + "raw_apical_rms": 0.07018123055776246, + "teaching_rms": 0.0008352924901350643, + "traffic_rms": 0.07017297314327207 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.3171905161898394e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5286738991683036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011895542390380374, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006125055885314941 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139103949069977, + "mean_feedback_forward_cosine": 0.9997563688985763, + "mean_feedback_forward_relative_error": 0.021637974652431665, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009133005875981959, + "instruction_rms": 0.000913300587630574, + "raw_apical_rms": 0.07017973000798261, + "teaching_rms": 0.0009133005875981959, + "traffic_rms": 0.07016999293303874 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.107489187139436e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5244305160354308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011809509510316457, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006561208536401391 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139101341366768, + "mean_feedback_forward_cosine": 0.9997563631303849, + "mean_feedback_forward_relative_error": 0.021637904157321298, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000827888955229611, + "instruction_rms": 0.0008278889548516962, + "raw_apical_rms": 0.07018139733869468, + "teaching_rms": 0.000827888955229611, + "traffic_rms": 0.07017350264932556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.5706521234478714e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527365136418755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011836103361401371, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006097571739032865 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139098361134529, + "mean_feedback_forward_cosine": 0.9997563650531154, + "mean_feedback_forward_relative_error": 0.021637833076378992, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008271057348685357, + "instruction_rms": 0.0008271057344927103, + "raw_apical_rms": 0.0701830064866668, + "teaching_rms": 0.0008271057348685357, + "traffic_rms": 0.07017452408400462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 3.791843283806884e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529119601998768e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01172571953440026, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.005912614208087325 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313909612596035, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.02163776025296219, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008146562174710621, + "instruction_rms": 0.0008146562172617252, + "raw_apical_rms": 0.07017955391646784, + "teaching_rms": 0.0008146562174710621, + "traffic_rms": 0.07017148572840433 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.144100174130472e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5312665939050198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011699781173596043, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.005879201124161482 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139093518257141, + "mean_feedback_forward_cosine": 0.9997563535167325, + "mean_feedback_forward_relative_error": 0.021637686858734777, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000831010217548728, + "instruction_rms": 0.0008310102176497575, + "raw_apical_rms": 0.07018308351432845, + "teaching_rms": 0.000831010217548728, + "traffic_rms": 0.07017445661334938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.759110794338994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5292206490358337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012051798098408104, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006030674913972616 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139090910553932, + "mean_feedback_forward_cosine": 0.9997563785122286, + "mean_feedback_forward_relative_error": 0.021637615432301836, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008653160871799459, + "instruction_rms": 0.0008653160870950886, + "raw_apical_rms": 0.0701803379876294, + "teaching_rms": 0.0008653160871799459, + "traffic_rms": 0.07017148296286081 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.355867844196837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5270904337238152e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012019349855465829, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006269852671995759 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139089047908783, + "mean_feedback_forward_cosine": 0.9997563727440373, + "mean_feedback_forward_relative_error": 0.0216375419329251, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008638587831392026, + "instruction_rms": 0.0008638587827622599, + "raw_apical_rms": 0.07017988164308915, + "teaching_rms": 0.0008638587831392026, + "traffic_rms": 0.07017073324045697 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.622512150831426e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5253112846296542e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01190656266396917, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006076337253674865 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139086440205574, + "mean_feedback_forward_cosine": 0.9997563727440373, + "mean_feedback_forward_relative_error": 0.021637468012951074, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008991722388024863, + "instruction_rms": 0.0008991722390854223, + "raw_apical_rms": 0.07017635572476152, + "teaching_rms": 0.0008991722388024863, + "traffic_rms": 0.07016603917640671 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.997853153554388e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.530415609004951e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011988472336836099, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0066249953493475915 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139083832502365, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.021637392936334494, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008324163210763149, + "instruction_rms": 0.0008324163210412614, + "raw_apical_rms": 0.07017822832269285, + "teaching_rms": 0.0008324163210763149, + "traffic_rms": 0.07016959894258296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.642785794179872e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5290053005521696e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011841474660883746, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.005963247852548957 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139080852270126, + "mean_feedback_forward_cosine": 0.9997563708213068, + "mean_feedback_forward_relative_error": 0.02163732009789636, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008975362948856293, + "instruction_rms": 0.0008975362949277453, + "raw_apical_rms": 0.07017975651367861, + "teaching_rms": 0.0008975362948856293, + "traffic_rms": 0.07016992034967445 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.265225766493131e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5288708603040966e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01182780420983723, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.006455110897123814 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139078989624977, + "mean_feedback_forward_cosine": 0.9997563804349592, + "mean_feedback_forward_relative_error": 0.021637246117837006, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008255834192435847, + "instruction_rms": 0.0008255834195698017, + "raw_apical_rms": 0.07017310714224047, + "teaching_rms": 0.0008255834192435847, + "traffic_rms": 0.0701651303002121 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.364056493955104e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5308435967627553e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011877196849235256, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.005828254036456346 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139077126979828, + "mean_feedback_forward_cosine": 0.9997563804349592, + "mean_feedback_forward_relative_error": 0.021637174751489394, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008587196618536937, + "instruction_rms": 0.0008587196615387869, + "raw_apical_rms": 0.07017349567438183, + "teaching_rms": 0.0008587196618536937, + "traffic_rms": 0.07016435416098747 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 4.398664898985771e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5317104033707283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012031463454533913, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.006170055857598782 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139074146747589, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.021637101582581957, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008308808167954374, + "instruction_rms": 0.0008308808168024899, + "raw_apical_rms": 0.07017087373690241, + "teaching_rms": 0.0008308808167954374, + "traffic_rms": 0.0701620163000156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631088018417, + "maximum_absolute_post_projection_soma_slope": 5.509829037464442e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5282689848842672e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011796754806936496, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.005872546712607145 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313907228410244, + "mean_feedback_forward_cosine": 0.9997563765894982, + "mean_feedback_forward_relative_error": 0.021637029164741115, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008117500631573633, + "instruction_rms": 0.0008117500631777582, + "raw_apical_rms": 0.07017044580576692, + "teaching_rms": 0.0008117500631573633, + "traffic_rms": 0.07016250237034652 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 3.76164344118024e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5307327171691512e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011852708223547039, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.00592443825788796 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139070048928261, + "mean_feedback_forward_cosine": 0.9997563669758458, + "mean_feedback_forward_relative_error": 0.02163696038206258, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000828283373564146, + "instruction_rms": 0.0008282833737996556, + "raw_apical_rms": 0.07017043938324789, + "teaching_rms": 0.000828283373564146, + "traffic_rms": 0.07016185720467448 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225631833076477, + "maximum_absolute_post_projection_soma_slope": 4.8528104201750466e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5308556307284527e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01183360771782816, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005726474728733301 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031390681862831116, + "mean_feedback_forward_cosine": 0.9997563823576896, + "mean_feedback_forward_relative_error": 0.021636890292528174, + "min_feedback_forward_cosine": 0.9995083808898926 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008030064494273148, + "instruction_rms": 0.0008030064493525704, + "raw_apical_rms": 0.07016814097972814, + "teaching_rms": 0.0008030064494273148, + "traffic_rms": 0.07015999214687181 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11225630342960358, + "maximum_absolute_post_projection_soma_slope": 4.896140026744433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5266947256581533e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011849112651321754, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0055932730484753845 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9226, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.36560716552734374, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2529336320, + "peak_memory_reserved_bytes": 3821010944, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 4.726146940470244e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 3.1396520938642425e-08, + "min_residual_soma_slope": -4.726146940470244e-08, + "mse": 3.4760263535321834e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.180593616172365e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.180593616172365e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1373100280761719, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.041075944900512695, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 14783.157014369965, + "train_wall_s": 14779.435670375824 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.01729196310043335, + 0.01262203324586153, + 0.008211836218833923, + 0.007364083081483841, + 0.005412328988313675, + 0.0057131508365273476, + 0.0037188821006566286, + 0.00471151527017355, + 0.0028882285114377737, + 0.0036457546520978212, + 0.002361578866839409, + 0.005046239122748375, + 0.003541539190337062, + 0.004347427748143673, + 0.002827239455655217, + 0.003586750477552414, + 0.0024223714135587215, + 0.003299170872196555, + 0.0020495131611824036, + 0.0028001274913549423, + 0.0017875225748866796, + 0.0038567481096833944, + 0.002763295779004693, + 0.0034892389085143805, + 0.0024175953585654497, + 0.0032130118925124407, + 0.002143308985978365, + 0.0028988828416913748, + 0.001982533372938633, + 0.002512215403839946, + 0.0019311783835291862 + ], + "realized_traffic_instruction_rms_ratio": [ + 3.999999523162842, + 4.0, + 3.999999523162842, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.000000476837158, + 3.999999523162842, + 3.999999761581421, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 3.999999761581421, + 3.999999761581421, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.0916287824511528, + 0.06876388937234879, + 0.02820105291903019, + 0.03799518570303917, + 0.01515876967459917, + 0.030091680586338043, + 0.008624541573226452, + 0.024622133001685143, + 0.005851417779922485, + 0.019561035558581352, + 0.004281718749552965, + 0.02729577012360096, + 0.007838867604732513, + 0.022911302745342255, + 0.0055216592736542225, + 0.018870050087571144, + 0.0043528093956410885, + 0.018290789797902107, + 0.003409921657294035, + 0.015034525655210018, + 0.0027709321584552526, + 0.020493168383836746, + 0.005571093410253525, + 0.018566109240055084, + 0.004393921233713627, + 0.016968388110399246, + 0.0035954569466412067, + 0.015505515038967133, + 0.003123497124761343, + 0.013197866268455982, + 0.002851189812645316 + ], + "unscaled_traffic_rms": [ + 0.7548703551292419, + 0.7342244982719421, + 1.164755940437317, + 0.7752648591995239, + 1.4281710386276245, + 0.7594326138496399, + 1.7247905731201172, + 0.765411376953125, + 1.9743785858154297, + 0.7455136179924011, + 2.2061970233917236, + 0.739490270614624, + 1.8071688413619995, + 0.7590014338493347, + 2.0481085777282715, + 0.7603054642677307, + 2.2260303497314453, + 0.7214933633804321, + 2.4041762351989746, + 0.7449859380722046, + 2.5803916454315186, + 0.7527871131896973, + 1.9840240478515625, + 0.7517436742782593, + 2.200854539871216, + 0.7574111819267273, + 2.3844635486602783, + 0.747832715511322, + 2.5388636589050293, + 0.7614004611968994, + 2.7092947959899902 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684206618894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 72745580756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 2745665106100224 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json new file mode 100644 index 0000000..c49faa1 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json @@ -0,0 +1,6861 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5012, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9996383130550385, + "feedback_forward_cosine": [ + 0.9997127652168274, + 0.9996895790100098, + 0.9996905326843262, + 0.9996283054351807, + 0.999725878238678, + 0.999717116355896, + 0.9997005462646484, + 0.9997110366821289, + 0.9997404217720032, + 0.9996763467788696, + 0.9998053908348083, + 0.9998557567596436, + 0.9997575879096985, + 0.9997172355651855, + 0.9997953772544861, + 0.9997363686561584, + 0.9997872114181519, + 0.9997088313102722, + 0.9997762441635132, + 0.9996641874313354, + 0.9997904896736145, + 0.9998691082000732, + 0.999842643737793, + 0.9998258948326111, + 0.9998345375061035, + 0.9997801780700684, + 0.9998278617858887, + 0.999695897102356, + 0.999720573425293, + 0.9994416236877441, + 0.9999860525131226 + ], + "feedback_forward_norm_ratio": [ + 0.9994192123413086, + 1.00136399269104, + 0.9990062713623047, + 1.001321792602539, + 0.998970091342926, + 1.0013912916183472, + 0.9995063543319702, + 1.001404047012329, + 0.999437689781189, + 1.001446008682251, + 0.9992797374725342, + 1.0001482963562012, + 0.9997099041938782, + 1.0010145902633667, + 0.9991261959075928, + 1.0012949705123901, + 0.9994481801986694, + 1.00155770778656, + 0.9994058609008789, + 1.0015500783920288, + 0.9997960925102234, + 1.0003494024276733, + 0.9996151328086853, + 1.0006310939788818, + 0.9999629259109497, + 1.000908613204956, + 0.9997559189796448, + 1.0009875297546387, + 0.9993951916694641, + 1.0011926889419556, + 1.002600073814392 + ], + "feedback_forward_relative_error": [ + 0.02397162653505802, + 0.024972213432192802, + 0.024884255602955818, + 0.02731448970735073, + 0.02342882938683033, + 0.02384459786117077, + 0.02446872368454933, + 0.024096522480249405, + 0.022784799337387085, + 0.025500789284706116, + 0.019736096262931824, + 0.016983557492494583, + 0.022016946226358414, + 0.02381383255124092, + 0.020233269780874252, + 0.023010481148958206, + 0.020631276071071625, + 0.024198558181524277, + 0.021154936403036118, + 0.02597862482070923, + 0.020470989868044853, + 0.016186153516173363, + 0.01774236187338829, + 0.01867579109966755, + 0.018197201192378998, + 0.020998291671276093, + 0.01855279877781868, + 0.024693824350833893, + 0.023642484098672867, + 0.033460669219493866, + 0.005887217819690704 + ], + "innovation_negative_gradient_cosine": [ + 0.9996783137321472, + 0.9995768070220947, + 0.9996729493141174, + 0.9996098875999451, + 0.9996705055236816, + 0.999569833278656, + 0.9996834993362427, + 0.9995659589767456, + 0.9996894001960754, + 0.9996659755706787, + 0.9997035264968872, + 0.9997659921646118, + 0.9997296333312988, + 0.9996403455734253, + 0.9997360110282898, + 0.9996874332427979, + 0.9997498989105225, + 0.9996836185455322, + 0.9997683763504028, + 0.9997152090072632, + 0.9997637271881104, + 0.9998130202293396, + 0.999793529510498, + 0.9997916221618652, + 0.9998072981834412, + 0.9998267889022827, + 0.9998589754104614, + 0.999876081943512, + 0.9999130368232727, + 0.9999557733535767, + 0.7652151584625244 + ], + "instruction_negative_gradient_cosine": [ + 0.999678373336792, + 0.9995768070220947, + 0.9996729493141174, + 0.9996098279953003, + 0.9996705055236816, + 0.999569833278656, + 0.9996834993362427, + 0.9995660185813904, + 0.9996894001960754, + 0.9996659755706787, + 0.9997035264968872, + 0.9997659921646118, + 0.9997296929359436, + 0.9996403455734253, + 0.9997360706329346, + 0.9996874332427979, + 0.9997498989105225, + 0.9996836185455322, + 0.9997683763504028, + 0.999715268611908, + 0.9997637271881104, + 0.9998130202293396, + 0.999793529510498, + 0.9997916221618652, + 0.9998073577880859, + 0.9998267889022827, + 0.9998589754104614, + 0.999876081943512, + 0.9999130368232727, + 0.9999557733535767, + 0.8075519800186157 + ], + "matched_negative_gradient_cosine": [ + 0.017511021345853806, + 0.03619465231895447, + 0.016825692728161812, + 0.005765100941061974, + 0.02352992631494999, + 0.044482000172138214, + -0.010311171412467957, + 0.011017613112926483, + -0.02334025502204895, + 0.030446156859397888, + -0.044917747378349304, + -0.018352026119828224, + -0.04602780193090439, + 0.06982018053531647, + -0.05281446501612663, + -0.017713584005832672, + -0.054920852184295654, + 0.07297352701425552, + -0.08952753245830536, + 0.09800183773040771, + -0.07944676280021667, + -0.020444614812731743, + -0.07381309568881989, + 0.075150266289711, + -0.043960269540548325, + 0.10424681007862091, + -0.05456995964050293, + 0.08944201469421387, + -0.008397681638598442, + -0.0261323694139719, + 0.08102748543024063 + ], + "max_norm_match_direction_error": 2.384185791015625e-07, + "max_norm_match_relative_error": 1.668656892661602e-07, + "mean_feedback_forward_cosine": 0.9997487606540802, + "mean_feedback_forward_relative_error": 0.021984909991583517, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.11954792588949203, + "max_absolute_post_projection_soma_slope": 2.2193493265376674e-08, + "max_absolute_pre_projection_soma_slope": 0.11954792588949203, + "max_positive_post_projection_soma_slope": 2.2193493265376674e-08, + "min_post_projection_soma_slope": -1.9364406966815295e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.387651846589325e-08, + "pre_projection_traffic_rms_ratio": 0.020527995727009703 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.020527995726996012, + "raw_negative_gradient_cosine": [ + 0.017511021345853806, + 0.03619465231895447, + 0.016825692728161812, + 0.005765100475400686, + 0.02352992817759514, + 0.044482000172138214, + -0.010311171412467957, + 0.011017613112926483, + -0.0233402531594038, + 0.030446156859397888, + -0.044917747378349304, + -0.018352027982473373, + -0.04602780193090439, + 0.06982018053531647, + -0.05281446501612663, + -0.017713584005832672, + -0.054920852184295654, + 0.07297352701425552, + -0.08952753245830536, + 0.09800183773040771, + -0.07944676280021667, + -0.020444616675376892, + -0.07381309568881989, + 0.0751502737402916, + -0.04396027326583862, + 0.10424681007862091, + -0.05456995591521263, + 0.08944201469421387, + -0.008397682569921017, + -0.0261323694139719, + 0.08102747797966003 + ], + "teaching_negative_gradient_cosine": [ + 0.9996783137321472, + 0.9995768070220947, + 0.9996729493141174, + 0.9996098875999451, + 0.9996705055236816, + 0.999569833278656, + 0.9996834993362427, + 0.9995659589767456, + 0.9996894001960754, + 0.9996659755706787, + 0.9997035264968872, + 0.9997659921646118, + 0.9997296333312988, + 0.9996403455734253, + 0.9997360110282898, + 0.9996874332427979, + 0.9997498989105225, + 0.9996836185455322, + 0.9997683763504028, + 0.9997152090072632, + 0.9997637271881104, + 0.9998130202293396, + 0.999793529510498, + 0.9997916221618652, + 0.9998072981834412, + 0.9998267889022827, + 0.9998589754104614, + 0.999876081943512, + 0.9999130368232727, + 0.9999557733535767, + 0.7652151584625244 + ], + "traffic_instruction_rms_ratio": 80.49653491732126, + "used_negative_gradient_cosine": [ + 0.9996783137321472, + 0.9995768070220947, + 0.9996729493141174, + 0.9996098875999451, + 0.9996705055236816, + 0.999569833278656, + 0.9996834993362427, + 0.9995659589767456, + 0.9996894001960754, + 0.9996659755706787, + 0.9997035264968872, + 0.9997659921646118, + 0.9997296333312988, + 0.9996403455734253, + 0.9997360110282898, + 0.9996874332427979, + 0.9997498989105225, + 0.9996836185455322, + 0.9997683763504028, + 0.9997152090072632, + 0.9997637271881104, + 0.9998130202293396, + 0.999793529510498, + 0.9997916221618652, + 0.9998072981834412, + 0.9998267889022827, + 0.9998589754104614, + 0.999876081943512, + 0.9999130368232727, + 0.9999557733535767, + 0.7652151584625244 + ], + "wall_s": 0.19591236114501953 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.339983344078064, + "mean_feedback_forward_cosine": 0.285654223974674, + "mean_feedback_forward_relative_error": 1.1921536480226824, + "min_feedback_forward_cosine": 0.10241539031267166 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002080249224120902, + "instruction_rms": 0.0020802492241344113, + "raw_apical_rms": 0.040167524853474024, + "teaching_rms": 0.002080249224120902, + "traffic_rms": 0.040075922799139324 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.486498994538124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.008953587271665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018426994423974323, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.940824231262207 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3133119344711304, + "mean_feedback_forward_cosine": 0.38046929961250675, + "mean_feedback_forward_relative_error": 1.1108155231322012, + "min_feedback_forward_cosine": 0.13950735330581665 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026746308311653903, + "instruction_rms": 0.0026746308302078702, + "raw_apical_rms": 0.048090439225993986, + "teaching_rms": 0.0026746308311653903, + "traffic_rms": 0.0480152001535155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.2503241814983994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7662187264540357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02232825240469484, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.3879668685531616 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2808902263641357, + "mean_feedback_forward_cosine": 0.4578625694397957, + "mean_feedback_forward_relative_error": 1.0404118030301985, + "min_feedback_forward_cosine": 0.18380451202392578 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003284515411786943, + "instruction_rms": 0.0032845154118905534, + "raw_apical_rms": 0.05424790974002508, + "teaching_rms": 0.003284515411786943, + "traffic_rms": 0.05414754658142877 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.6566571992580066e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.716860835560102e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.025622003462195987, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0928825889587401 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2442213296890259, + "mean_feedback_forward_cosine": 0.5184986600952763, + "mean_feedback_forward_relative_error": 0.9816472491910381, + "min_feedback_forward_cosine": 0.23061507940292358 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035898430064823287, + "instruction_rms": 0.003589843006865477, + "raw_apical_rms": 0.05909211230081736, + "teaching_rms": 0.0035898430064823287, + "traffic_rms": 0.058982153565529306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.9844990478977707e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6608289930903792e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.026485848129165296, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8990384727287293 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2057592868804932, + "mean_feedback_forward_cosine": 0.5672492798297636, + "mean_feedback_forward_relative_error": 0.9311478887834856, + "min_feedback_forward_cosine": 0.2770645022392273 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036866174041923906, + "instruction_rms": 0.0036866174047952317, + "raw_apical_rms": 0.062393623430155365, + "teaching_rms": 0.0036866174041923906, + "traffic_rms": 0.0622836793304873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4490208971883476e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.627344198930493e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.026418282037462514, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7792262822723389 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1737024784088135, + "mean_feedback_forward_cosine": 0.6083217032494084, + "mean_feedback_forward_relative_error": 0.8856505828519021, + "min_feedback_forward_cosine": 0.3189030885696411 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037864097929684636, + "instruction_rms": 0.0037864097933226737, + "raw_apical_rms": 0.06482770913559593, + "teaching_rms": 0.0037864097929684636, + "traffic_rms": 0.06471622521698416 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.9914069443748303e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6099970994611725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02584536381721676, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7007547276115418 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1478946208953857, + "mean_feedback_forward_cosine": 0.644517679368296, + "mean_feedback_forward_relative_error": 0.8437995641462265, + "min_feedback_forward_cosine": 0.3569311797618866 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003814658322369643, + "instruction_rms": 0.0038146583232081316, + "raw_apical_rms": 0.06741707480641723, + "teaching_rms": 0.003814658322369643, + "traffic_rms": 0.06730804348495129 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.5306700734972765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6033405663510874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.026268299145706272, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6401975685310364 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1219756603240967, + "mean_feedback_forward_cosine": 0.6752194012365034, + "mean_feedback_forward_relative_error": 0.8063726867398908, + "min_feedback_forward_cosine": 0.3872883915901184 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003779530891957654, + "instruction_rms": 0.003779530892228306, + "raw_apical_rms": 0.06958201970921558, + "teaching_rms": 0.003779530891957654, + "traffic_rms": 0.06947828958060741 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 3.070394427595602e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6003344614258366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.024182602444106506, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5894514981269836 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0964499711990356, + "mean_feedback_forward_cosine": 0.7017301407552534, + "mean_feedback_forward_relative_error": 0.7723443863853332, + "min_feedback_forward_cosine": 0.4165016710758209 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037507535715463397, + "instruction_rms": 0.003750753572380765, + "raw_apical_rms": 0.07080815978236496, + "teaching_rms": 0.0037507535715463397, + "traffic_rms": 0.07070753242282429 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.2875282112977402e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5890318169785002e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.025065987759960773, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5456703003311157 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.070331335067749, + "mean_feedback_forward_cosine": 0.7265749477571056, + "mean_feedback_forward_relative_error": 0.7393224921918684, + "min_feedback_forward_cosine": 0.4450567662715912 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003801102928502972, + "instruction_rms": 0.0038011029290939795, + "raw_apical_rms": 0.07245479716633492, + "teaching_rms": 0.003801102928502972, + "traffic_rms": 0.07235376710914479 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.225028066822233e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5793764682374934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02560264079951687, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5172618599033356 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.044991135597229, + "mean_feedback_forward_cosine": 0.7481149896498649, + "mean_feedback_forward_relative_error": 0.7090476962827867, + "min_feedback_forward_cosine": 0.47188228368759155 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003800919419561118, + "instruction_rms": 0.003800919419834835, + "raw_apical_rms": 0.07366327193167098, + "teaching_rms": 0.003800919419561118, + "traffic_rms": 0.07356389856252486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.0076314772031765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5778947745267913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.023600138084264186, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4883759662055969 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0188921689987183, + "mean_feedback_forward_cosine": 0.7675743420277873, + "mean_feedback_forward_relative_error": 0.6807664450137846, + "min_feedback_forward_cosine": 0.4993303716182709 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038081847436252703, + "instruction_rms": 0.0038081847428805735, + "raw_apical_rms": 0.07483873834969192, + "teaching_rms": 0.0038081847436252703, + "traffic_rms": 0.07474069233291468 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.383430424719336e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563779796097362e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.023511291375287888, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.46503282198905943 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9933239817619324, + "mean_feedback_forward_cosine": 0.7850865721702576, + "mean_feedback_forward_relative_error": 0.6539644872949969, + "min_feedback_forward_cosine": 0.5248597264289856 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037975004780219064, + "instruction_rms": 0.0037975004777826777, + "raw_apical_rms": 0.0751567703844696, + "teaching_rms": 0.0037975004780219064, + "traffic_rms": 0.07505935765865411 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.5978653894753734e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568034641812459e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.025134926309358284, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.44063143260002136 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9673856496810913, + "mean_feedback_forward_cosine": 0.8014208097611705, + "mean_feedback_forward_relative_error": 0.6281607074122275, + "min_feedback_forward_cosine": 0.5494941473007202 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038286869929528345, + "instruction_rms": 0.0038286869928880426, + "raw_apical_rms": 0.0755839140671827, + "teaching_rms": 0.0038286869929528345, + "traffic_rms": 0.07548557374813313 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.2704825514429103e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567613383472839e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02341548672107511, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.424243131275177 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.942949116230011, + "mean_feedback_forward_cosine": 0.8160869921407392, + "mean_feedback_forward_relative_error": 0.60416255506777, + "min_feedback_forward_cosine": 0.5720338821411133 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003815301638750463, + "instruction_rms": 0.003815301638505104, + "raw_apical_rms": 0.07630348870986853, + "teaching_rms": 0.003815301638750463, + "traffic_rms": 0.07620670851335472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 2.7082339926209897e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.579548432999604e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.023191719646902132, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.40831927920341493 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9173076152801514, + "mean_feedback_forward_cosine": 0.8298169855148562, + "mean_feedback_forward_relative_error": 0.5807071794425288, + "min_feedback_forward_cosine": 0.5949041247367859 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003888690006762004, + "instruction_rms": 0.0038886900062056274, + "raw_apical_rms": 0.07705134584979244, + "teaching_rms": 0.003888690006762004, + "traffic_rms": 0.0769517401391886 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 2.7501505073246335e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5709947772725082e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02365437562225305, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39128387632369993 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8924576640129089, + "mean_feedback_forward_cosine": 0.8425317406654358, + "mean_feedback_forward_relative_error": 0.5581389670410464, + "min_feedback_forward_cosine": 0.6168732643127441 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038875513240429053, + "instruction_rms": 0.0038875513250331934, + "raw_apical_rms": 0.0774914690791847, + "teaching_rms": 0.0038875513240429053, + "traffic_rms": 0.07739251392587451 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4438445933583353e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670128446355948e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02437334659023546, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3801311800289154 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8670241832733154, + "mean_feedback_forward_cosine": 0.8543807056642347, + "mean_feedback_forward_relative_error": 0.5363450382025011, + "min_feedback_forward_cosine": 0.6383365392684937 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00384500775541852, + "instruction_rms": 0.003845007755000397, + "raw_apical_rms": 0.07832073428570242, + "teaching_rms": 0.00384500775541852, + "traffic_rms": 0.07822466862381727 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.5322519192627624e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5564336764432264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02188136392313046, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3651800305223465 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8419966101646423, + "mean_feedback_forward_cosine": 0.865056985808957, + "mean_feedback_forward_relative_error": 0.5159447313316406, + "min_feedback_forward_cosine": 0.6593444347381592 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003845568267843205, + "instruction_rms": 0.0038455682693480553, + "raw_apical_rms": 0.07872827558104348, + "teaching_rms": 0.003845568267843205, + "traffic_rms": 0.07863286559487295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 2.6638568684234087e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.552634404029924e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022451397270068573, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.35390729375839236 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8158823847770691, + "mean_feedback_forward_cosine": 0.875196099281311, + "mean_feedback_forward_relative_error": 0.49568016106082546, + "min_feedback_forward_cosine": 0.6797770857810974 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038774594622224185, + "instruction_rms": 0.0038774594629651968, + "raw_apical_rms": 0.07898112474063972, + "teaching_rms": 0.0038774594622224185, + "traffic_rms": 0.07888394814688401 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.474346416647677e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558705973092894e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022742344210804517, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.34763749789237974 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7906211614608765, + "mean_feedback_forward_cosine": 0.8843311071395874, + "mean_feedback_forward_relative_error": 0.4769887044545143, + "min_feedback_forward_cosine": 0.6992702484130859 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003802348199810035, + "instruction_rms": 0.003802348199231994, + "raw_apical_rms": 0.07967834579403497, + "teaching_rms": 0.003802348199810035, + "traffic_rms": 0.07958576632750877 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 2.8695327003447346e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5541955505899907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0225150335205604, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3351114765357971 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7663038372993469, + "mean_feedback_forward_cosine": 0.8926228957791482, + "mean_feedback_forward_relative_error": 0.45923874166703993, + "min_feedback_forward_cosine": 0.7173070907592773 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003746605771084313, + "instruction_rms": 0.003746605771660771, + "raw_apical_rms": 0.07995686548010453, + "teaching_rms": 0.003746605771084313, + "traffic_rms": 0.07986741843977396 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.128937819951716e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5571558664184326e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02191969048439542, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32765958612442014 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7423281669616699, + "mean_feedback_forward_cosine": 0.9003835993428384, + "mean_feedback_forward_relative_error": 0.4420319882131392, + "min_feedback_forward_cosine": 0.7346341609954834 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003803798781482709, + "instruction_rms": 0.0038037987819786482, + "raw_apical_rms": 0.0802889870233249, + "teaching_rms": 0.003803798781482709, + "traffic_rms": 0.08019723992911532 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.6047445089716348e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5555373872400654e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021384591420741537, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3194572360420227 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7184553742408752, + "mean_feedback_forward_cosine": 0.9075927638238476, + "mean_feedback_forward_relative_error": 0.42546406532487563, + "min_feedback_forward_cosine": 0.7514128684997559 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003811076587604591, + "instruction_rms": 0.003811076587067008, + "raw_apical_rms": 0.08061216923055259, + "teaching_rms": 0.003811076587604591, + "traffic_rms": 0.0805201696573842 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 2.9550477620432503e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5568044288401618e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02247422627232819, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.30688063206672667 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6946475505828857, + "mean_feedback_forward_cosine": 0.9145379758650257, + "mean_feedback_forward_relative_error": 0.4089530086325061, + "min_feedback_forward_cosine": 0.7675841450691223 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00382989182811828, + "instruction_rms": 0.0038298918277047014, + "raw_apical_rms": 0.08083736719136075, + "teaching_rms": 0.00382989182811828, + "traffic_rms": 0.08074476792836242 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.423254308325795e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556297327911852e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02161529452935168, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.30793094880104066 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6719034910202026, + "mean_feedback_forward_cosine": 0.9207627504102646, + "mean_feedback_forward_relative_error": 0.39359850796961016, + "min_feedback_forward_cosine": 0.7825264930725098 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037957908672480263, + "instruction_rms": 0.003795790868013067, + "raw_apical_rms": 0.08113030517176986, + "teaching_rms": 0.0037957908672480263, + "traffic_rms": 0.08103955773970174 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.9971225501412846e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563857698410766e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022212461071887117, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2988512475013733 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.648636519908905, + "mean_feedback_forward_cosine": 0.9265817922930564, + "mean_feedback_forward_relative_error": 0.3786434334131979, + "min_feedback_forward_cosine": 0.7969982028007507 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003789280125772157, + "instruction_rms": 0.003789280125846186, + "raw_apical_rms": 0.08137693864197929, + "teaching_rms": 0.003789280125772157, + "traffic_rms": 0.08128687373600725 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.5351011007378474e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547528127230909e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022590934673623334, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.2955006490373611 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6280379295349121, + "mean_feedback_forward_cosine": 0.9317853719957413, + "mean_feedback_forward_relative_error": 0.36478087161817857, + "min_feedback_forward_cosine": 0.8096004724502563 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003720107890496055, + "instruction_rms": 0.0037201078904678503, + "raw_apical_rms": 0.08158154932965317, + "teaching_rms": 0.003720107890496055, + "traffic_rms": 0.0814947290184295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.66785775693279e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5439884771976598e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02249486728222325, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28047562375068663 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.606637716293335, + "mean_feedback_forward_cosine": 0.9367853633819088, + "mean_feedback_forward_relative_error": 0.3509458897094573, + "min_feedback_forward_cosine": 0.8222372531890869 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003773844691080946, + "instruction_rms": 0.003773844690461671, + "raw_apical_rms": 0.08196960446469068, + "teaching_rms": 0.003773844691080946, + "traffic_rms": 0.08188111289104735 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.5396712288738854e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5454621360364765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021035724662669116, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2748355854034424 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5845741033554077, + "mean_feedback_forward_cosine": 0.941536255421177, + "mean_feedback_forward_relative_error": 0.3373694977452678, + "min_feedback_forward_cosine": 0.8346878290176392 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038041466624678063, + "instruction_rms": 0.0038041466607106624, + "raw_apical_rms": 0.08262488422964115, + "teaching_rms": 0.0038041466624678063, + "traffic_rms": 0.08253555321885514 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.025534311973388e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5504034505469517e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021575066880123162, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2817251724815369 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5648671984672546, + "mean_feedback_forward_cosine": 0.9457164791322523, + "mean_feedback_forward_relative_error": 0.32496031229534456, + "min_feedback_forward_cosine": 0.8454329967498779 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003696997936087001, + "instruction_rms": 0.0036969979353304325, + "raw_apical_rms": 0.08311236722775739, + "teaching_rms": 0.003696997936087001, + "traffic_rms": 0.08302831970609409 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 2.9984740024247e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550059349709595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021562584337519015, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.2675520391702652 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5451713800430298, + "mean_feedback_forward_cosine": 0.9496755234656795, + "mean_feedback_forward_relative_error": 0.3127627603469356, + "min_feedback_forward_cosine": 0.8559471368789673 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003741085292343212, + "instruction_rms": 0.0037410852929531577, + "raw_apical_rms": 0.08343884742313849, + "teaching_rms": 0.003741085292343212, + "traffic_rms": 0.08335299450034313 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 2.6218115678489085e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539819986029742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022153264227316255, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26477368724822997 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5258288383483887, + "mean_feedback_forward_cosine": 0.9532831811135815, + "mean_feedback_forward_relative_error": 0.3011800134854932, + "min_feedback_forward_cosine": 0.8658452033996582 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036989109826581186, + "instruction_rms": 0.00369891098163216, + "raw_apical_rms": 0.08392184744006817, + "teaching_rms": 0.0036989109826581186, + "traffic_rms": 0.08383847547201535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.780694252635385e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543599796802316e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021519108977254647, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.25460036865234376 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5055931210517883, + "mean_feedback_forward_cosine": 0.9568271175507577, + "mean_feedback_forward_relative_error": 0.28946466191161063, + "min_feedback_forward_cosine": 0.8758914470672607 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003766140403572987, + "instruction_rms": 0.0037661404038988712, + "raw_apical_rms": 0.08434165028912659, + "teaching_rms": 0.003766140403572987, + "traffic_rms": 0.08425549402353745 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.603221638253217e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5315106253461986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021377326771947168, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2593456088924408 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4867854714393616, + "mean_feedback_forward_cosine": 0.9599301218986511, + "mean_feedback_forward_relative_error": 0.278771095218197, + "min_feedback_forward_cosine": 0.8847917914390564 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003660920335533748, + "instruction_rms": 0.0036609203352707892, + "raw_apical_rms": 0.08445576094726165, + "teaching_rms": 0.003660920335533748, + "traffic_rms": 0.084374537128479 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4792990771516088e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547704320857565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020858484746963266, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2517908228492737 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4684971868991852, + "mean_feedback_forward_cosine": 0.9628444890822133, + "mean_feedback_forward_relative_error": 0.26838447081465877, + "min_feedback_forward_cosine": 0.8932062387466431 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003676764120386712, + "instruction_rms": 0.0036767641207572197, + "raw_apical_rms": 0.08501292971141325, + "teaching_rms": 0.003676764120386712, + "traffic_rms": 0.08493144475411925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.3919136893368886e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542121244602389e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02120021736212017, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25233236048698426 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45047035813331604, + "mean_feedback_forward_cosine": 0.9655597498339992, + "mean_feedback_forward_relative_error": 0.25830644297022975, + "min_feedback_forward_cosine": 0.9012092351913452 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003679990967941737, + "instruction_rms": 0.0036799909676135684, + "raw_apical_rms": 0.0851496850663049, + "teaching_rms": 0.003679990967941737, + "traffic_rms": 0.08506827021794751 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.9648782984281752e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539612328170669e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020773911571546892, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2458469243335724 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4340652823448181, + "mean_feedback_forward_cosine": 0.9680396472254107, + "mean_feedback_forward_relative_error": 0.24874176445507235, + "min_feedback_forward_cosine": 0.9081953763961792 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003644402866019241, + "instruction_rms": 0.0036444028657358156, + "raw_apical_rms": 0.08496649589575904, + "teaching_rms": 0.003644402866019241, + "traffic_rms": 0.08488620220215817 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.3739916699128116e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5451101029528e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021873115790655737, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.23864479013442993 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4173719584941864, + "mean_feedback_forward_cosine": 0.9703690621160692, + "mean_feedback_forward_relative_error": 0.23945583859758993, + "min_feedback_forward_cosine": 0.9150651693344116 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003640872765862737, + "instruction_rms": 0.0036408727650291996, + "raw_apical_rms": 0.08512538467006531, + "teaching_rms": 0.003640872765862737, + "traffic_rms": 0.0850452255702441 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4701588330344748e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5419072895932438e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020875260569160312, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23881911232948302 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4010053873062134, + "mean_feedback_forward_cosine": 0.9725564218336537, + "mean_feedback_forward_relative_error": 0.23038030391739261, + "min_feedback_forward_cosine": 0.9214887022972107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003635837633583387, + "instruction_rms": 0.0036358376332435707, + "raw_apical_rms": 0.08567487952189785, + "teaching_rms": 0.003635837633583387, + "traffic_rms": 0.08559602130822146 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.317799368574015e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5412768536771445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02014687900955798, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.24160078894138337 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3865017294883728, + "mean_feedback_forward_cosine": 0.9744961165612743, + "mean_feedback_forward_relative_error": 0.22201711299919313, + "min_feedback_forward_cosine": 0.9270517826080322 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003607355159570028, + "instruction_rms": 0.003607355159732388, + "raw_apical_rms": 0.0859003992734445, + "teaching_rms": 0.003607355159570028, + "traffic_rms": 0.08582259692315432 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.3608007992947933e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.533301772029077e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020296927779150903, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22934025659561158 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37123191356658936, + "mean_feedback_forward_cosine": 0.9763900618399343, + "mean_feedback_forward_relative_error": 0.21357049264254108, + "min_feedback_forward_cosine": 0.9326528906822205 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036277093888634716, + "instruction_rms": 0.0036277093895101292, + "raw_apical_rms": 0.08627876412960948, + "teaching_rms": 0.0036277093888634716, + "traffic_rms": 0.08620047067783435 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.59356323084603e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.540223644719879e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020772425150943352, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23243314829826356 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35625848174095154, + "mean_feedback_forward_cosine": 0.9781526884725017, + "mean_feedback_forward_relative_error": 0.20543468070607032, + "min_feedback_forward_cosine": 0.9379054307937622 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036174161382302055, + "instruction_rms": 0.003617416137268926, + "raw_apical_rms": 0.08653085731511036, + "teaching_rms": 0.0036174161382302055, + "traffic_rms": 0.08645306446533436 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.9466003676789114e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5411118523938734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019880979082758023, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22913747826576233 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3421632647514343, + "mean_feedback_forward_cosine": 0.9797359512698266, + "mean_feedback_forward_relative_error": 0.1978166350914586, + "min_feedback_forward_cosine": 0.9427322745323181 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003573222538527244, + "instruction_rms": 0.003573222539230401, + "raw_apical_rms": 0.08665976544015823, + "teaching_rms": 0.003573222538527244, + "traffic_rms": 0.08658395129540997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4065903048153814e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538551983105687e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020373166720702253, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.2236249598789215 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3285570442676544, + "mean_feedback_forward_cosine": 0.9812291649080092, + "mean_feedback_forward_relative_error": 0.1903552626890521, + "min_feedback_forward_cosine": 0.9471590518951416 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036359997763206575, + "instruction_rms": 0.003635999776054257, + "raw_apical_rms": 0.08668425385848394, + "teaching_rms": 0.0036359997763206575, + "traffic_rms": 0.08660585540986522 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.697324674727497e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5412239550459958e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019691640759525866, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21998947902202606 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3161172568798065, + "mean_feedback_forward_cosine": 0.9826005139658528, + "mean_feedback_forward_relative_error": 0.1832409154984259, + "min_feedback_forward_cosine": 0.9510786533355713 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036017919184717483, + "instruction_rms": 0.0036017919177878965, + "raw_apical_rms": 0.08679553826776168, + "teaching_rms": 0.0036017919184717483, + "traffic_rms": 0.08671875062419489 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4171539436679268e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5493105717846082e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019258260655312977, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21795385506629944 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.303568571805954, + "mean_feedback_forward_cosine": 0.9839023101714349, + "mean_feedback_forward_relative_error": 0.1762339639807901, + "min_feedback_forward_cosine": 0.9548273682594299 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036224065081917217, + "instruction_rms": 0.0036224065075101566, + "raw_apical_rms": 0.08757546073479362, + "teaching_rms": 0.0036224065081917217, + "traffic_rms": 0.08749858906483558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4562153200236025e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5459714423101956e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01948794175861389, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2169464857006073 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29180771112442017, + "mean_feedback_forward_cosine": 0.9850710784235308, + "mean_feedback_forward_relative_error": 0.1696895678197184, + "min_feedback_forward_cosine": 0.9582520127296448 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035627054861949756, + "instruction_rms": 0.0035627054858939342, + "raw_apical_rms": 0.08753201662723528, + "teaching_rms": 0.0035627054861949756, + "traffic_rms": 0.08745727137465721 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.8947622965347364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5385903138064397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019584450661410183, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21340061833381652 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2798219919204712, + "mean_feedback_forward_cosine": 0.9861899172106097, + "mean_feedback_forward_relative_error": 0.16319163948778184, + "min_feedback_forward_cosine": 0.9615894556045532 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003586454904667169, + "instruction_rms": 0.0035864549048118025, + "raw_apical_rms": 0.08783464794161433, + "teaching_rms": 0.003586454904667169, + "traffic_rms": 0.0877592818811931 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.435359292007888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373971080316475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020086318690535118, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2142408071708679 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2687036693096161, + "mean_feedback_forward_cosine": 0.9871921731579688, + "mean_feedback_forward_relative_error": 0.15714425581597513, + "min_feedback_forward_cosine": 0.9645549654960632 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035468197410135076, + "instruction_rms": 0.003546819741902306, + "raw_apical_rms": 0.08781597742571949, + "teaching_rms": 0.0035468197410135076, + "traffic_rms": 0.08774253348408927 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.6614911163846955e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539736182174867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02008860360514407, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20817647980690002 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25753459334373474, + "mean_feedback_forward_cosine": 0.9881646960012375, + "mean_feedback_forward_relative_error": 0.15107314853418258, + "min_feedback_forward_cosine": 0.9674193859100342 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003591153247945613, + "instruction_rms": 0.0035911532483922353, + "raw_apical_rms": 0.08888751845762376, + "teaching_rms": 0.003591153247945613, + "traffic_rms": 0.08881303967673218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.5328708019856094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5296619512554614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020784637063976662, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.21359893213272094 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2474459558725357, + "mean_feedback_forward_cosine": 0.9890174731131522, + "mean_feedback_forward_relative_error": 0.14549297058293897, + "min_feedback_forward_cosine": 0.969895601272583 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003526777987799233, + "instruction_rms": 0.0035267779874768386, + "raw_apical_rms": 0.08914640316821017, + "teaching_rms": 0.003526777987799233, + "traffic_rms": 0.0890747284117013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 2.7976337690915898e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5399596731265748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020846874032681033, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20605351320266724 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23715411126613617, + "mean_feedback_forward_cosine": 0.9898492713128367, + "mean_feedback_forward_relative_error": 0.13989623443734261, + "min_feedback_forward_cosine": 0.9723339080810547 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003563383861972153, + "instruction_rms": 0.0035633838618865017, + "raw_apical_rms": 0.08945401136580772, + "teaching_rms": 0.003563383861972153, + "traffic_rms": 0.08938093314463431 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.3503424984028243e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5329162025305695e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02091112241956645, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20982917757987976 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22737513482570648, + "mean_feedback_forward_cosine": 0.9906215648497304, + "mean_feedback_forward_relative_error": 0.13446674616106094, + "min_feedback_forward_cosine": 0.9745501279830933 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035858712766025074, + "instruction_rms": 0.0035858712760856713, + "raw_apical_rms": 0.08941056224927504, + "teaching_rms": 0.0035858712766025074, + "traffic_rms": 0.08933648764502701 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.7436763022924424e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529668300633916e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020850600044323923, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.2077607313632965 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21776433289051056, + "mean_feedback_forward_cosine": 0.9913409717621342, + "mean_feedback_forward_relative_error": 0.1292032654727659, + "min_feedback_forward_cosine": 0.976641058921814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003615761091928174, + "instruction_rms": 0.0036157610921008503, + "raw_apical_rms": 0.08947213090545557, + "teaching_rms": 0.003615761091928174, + "traffic_rms": 0.08939706188408748 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.449490731417427e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5278470935355535e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02055420217669056, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.2050978511786461 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20912526547908783, + "mean_feedback_forward_cosine": 0.9919809314512438, + "mean_feedback_forward_relative_error": 0.12432593539837868, + "min_feedback_forward_cosine": 0.9784440398216248 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035560730274956016, + "instruction_rms": 0.003556073027054234, + "raw_apical_rms": 0.08932604240814324, + "teaching_rms": 0.0035560730274956016, + "traffic_rms": 0.08925308994395416 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.183000529816127e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529106656156594e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021170023524429888, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1972603488063812 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20059646666049957, + "mean_feedback_forward_cosine": 0.9925899640206368, + "mean_feedback_forward_relative_error": 0.1194925885046682, + "min_feedback_forward_cosine": 0.9801497459411621 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003561461384419712, + "instruction_rms": 0.003561461384152423, + "raw_apical_rms": 0.08969275877701628, + "teaching_rms": 0.003561461384419712, + "traffic_rms": 0.08961997947018287 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.3832802276710936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.527337583226723e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02007818182522912, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.199876754860878 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19231562316417694, + "mean_feedback_forward_cosine": 0.9931506206912379, + "mean_feedback_forward_relative_error": 0.11487585029775096, + "min_feedback_forward_cosine": 0.9817416667938232 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003572397417247057, + "instruction_rms": 0.0035723974179432344, + "raw_apical_rms": 0.09010630501835151, + "teaching_rms": 0.003572397417247057, + "traffic_rms": 0.09003342776220075 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.2090982194386015e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5342840484179384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01946575807985318, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.20152020866394044 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18480812013149261, + "mean_feedback_forward_cosine": 0.9936417764233004, + "mean_feedback_forward_relative_error": 0.11068257256861656, + "min_feedback_forward_cosine": 0.9831238389015198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035041394987490146, + "instruction_rms": 0.003504139498127927, + "raw_apical_rms": 0.09014100840569654, + "teaching_rms": 0.0035041394987490146, + "traffic_rms": 0.09007076417279689 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.389217522771105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5353137209224774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019352542806402275, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.19439666543006898 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17740245163440704, + "mean_feedback_forward_cosine": 0.99412194759615, + "mean_feedback_forward_relative_error": 0.10641636783557554, + "min_feedback_forward_cosine": 0.9844418168067932 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003526084237368824, + "instruction_rms": 0.003526084236286148, + "raw_apical_rms": 0.09088674860585018, + "teaching_rms": 0.003526084237368824, + "traffic_rms": 0.09081626399576186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.516608953508694e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5235592287471313e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019902515910294783, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19436994416236877 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17025071382522583, + "mean_feedback_forward_cosine": 0.994560380135813, + "mean_feedback_forward_relative_error": 0.1023753434419632, + "min_feedback_forward_cosine": 0.9856672286987305 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034963882295491245, + "instruction_rms": 0.0034963882292639507, + "raw_apical_rms": 0.09093808873264245, + "teaching_rms": 0.0034963882295491245, + "traffic_rms": 0.09086897345524017 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.537916626328297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.525981135991251e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019666085413254365, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18953499997138978 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1630403846502304, + "mean_feedback_forward_cosine": 0.9949728481231197, + "mean_feedback_forward_relative_error": 0.09843012438185754, + "min_feedback_forward_cosine": 0.986850380897522 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035419359508581994, + "instruction_rms": 0.00354193595166535, + "raw_apical_rms": 0.09101691362408658, + "teaching_rms": 0.0035419359508581994, + "traffic_rms": 0.09094582922106047 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.545998339404832e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5229900805215062e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019989621006640285, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19320359150886535 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15601927042007446, + "mean_feedback_forward_cosine": 0.9953637853745492, + "mean_feedback_forward_relative_error": 0.09453326667989453, + "min_feedback_forward_cosine": 0.9879511594772339 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035633617029206524, + "instruction_rms": 0.003563361703999385, + "raw_apical_rms": 0.09102371829446436, + "teaching_rms": 0.0035633617029206524, + "traffic_rms": 0.09095171219641386 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.389641716784354e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5265557477166158e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020356236874469368, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1960710004043579 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14947889745235443, + "mean_feedback_forward_cosine": 0.9957203307459431, + "mean_feedback_forward_relative_error": 0.0908293481315336, + "min_feedback_forward_cosine": 0.9889388084411621 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003560543210386606, + "instruction_rms": 0.0035605432112496387, + "raw_apical_rms": 0.09167314447405792, + "teaching_rms": 0.003560543210386606, + "traffic_rms": 0.09160188029210986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.232928318970153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5236609214455478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020269338767882495, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.19550201475143433 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14348924160003662, + "mean_feedback_forward_cosine": 0.9960384215078046, + "mean_feedback_forward_relative_error": 0.08738184293671962, + "min_feedback_forward_cosine": 0.9897992610931396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034696232870770043, + "instruction_rms": 0.0034696232865405376, + "raw_apical_rms": 0.09203038429619985, + "teaching_rms": 0.0034696232870770043, + "traffic_rms": 0.09196250854648237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.619381854809035e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529544492211799e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02043411407600841, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1877687189435959 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1377529799938202, + "mean_feedback_forward_cosine": 0.9963377137337962, + "mean_feedback_forward_relative_error": 0.08401348598061069, + "min_feedback_forward_cosine": 0.9905954599380493 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035635897063299986, + "instruction_rms": 0.0035635897062836677, + "raw_apical_rms": 0.09227704395719892, + "teaching_rms": 0.0035635897063299986, + "traffic_rms": 0.09220624337361381 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.530069475094933e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5475167122117337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020283477060844185, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18872840131759644 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13226574659347534, + "mean_feedback_forward_cosine": 0.996613863975771, + "mean_feedback_forward_relative_error": 0.0807801885229926, + "min_feedback_forward_cosine": 0.9913270473480225 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035459664115947555, + "instruction_rms": 0.003545966412260038, + "raw_apical_rms": 0.09278309030243255, + "teaching_rms": 0.0035459664115947555, + "traffic_rms": 0.09271337933639279 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.6564591233154715e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5287458128170955e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020796899204671513, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18797186532974244 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12683233618736267, + "mean_feedback_forward_cosine": 0.996869844775046, + "mean_feedback_forward_relative_error": 0.07766845966539075, + "min_feedback_forward_cosine": 0.9920245409011841 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034992089524370123, + "instruction_rms": 0.003499208952596316, + "raw_apical_rms": 0.09290394068064936, + "teaching_rms": 0.0034992089524370123, + "traffic_rms": 0.09283585880603441 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 3.064278075726179e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5227438798413417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018580219473450618, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1894784338283539 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12172558158636093, + "mean_feedback_forward_cosine": 0.9971017106886833, + "mean_feedback_forward_relative_error": 0.07473568606280512, + "min_feedback_forward_cosine": 0.9926533102989197 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034709234123134016, + "instruction_rms": 0.0034709234118668183, + "raw_apical_rms": 0.09275222339175077, + "teaching_rms": 0.0034709234123134016, + "traffic_rms": 0.09268501456842669 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.3856765330474445e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5249601275214636e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018866666273558735, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18372785980939865 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11663611978292465, + "mean_feedback_forward_cosine": 0.9973189061687838, + "mean_feedback_forward_relative_error": 0.07188035037008024, + "min_feedback_forward_cosine": 0.9932522177696228 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035141430347087653, + "instruction_rms": 0.0035141430339138617, + "raw_apical_rms": 0.09275169184950063, + "teaching_rms": 0.0035141430347087653, + "traffic_rms": 0.09268323340808805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.2397794547687226e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5226258692112242e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019234404414508882, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.18465503021240234 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11190527677536011, + "mean_feedback_forward_cosine": 0.9975191489342721, + "mean_feedback_forward_relative_error": 0.0691405072688095, + "min_feedback_forward_cosine": 0.993786096572876 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034866513598836857, + "instruction_rms": 0.003486651361087333, + "raw_apical_rms": 0.09323491465648073, + "teaching_rms": 0.0034866513598836857, + "traffic_rms": 0.09316780274526372 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.0512088733967175e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5213683826582096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019610186327961982, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18463716703891755 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10722721368074417, + "mean_feedback_forward_cosine": 0.9977076784256966, + "mean_feedback_forward_relative_error": 0.06646183962302823, + "min_feedback_forward_cosine": 0.9942919015884399 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00350020678437986, + "instruction_rms": 0.0035002067836061483, + "raw_apical_rms": 0.09346649483909661, + "teaching_rms": 0.00350020678437986, + "traffic_rms": 0.09339873528697154 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.745538007000505e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522508483052675e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019239783961154982, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.18354372583389283 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10300333797931671, + "mean_feedback_forward_cosine": 0.9978772421036998, + "mean_feedback_forward_relative_error": 0.06394916268125657, + "min_feedback_forward_cosine": 0.9947330951690674 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034676678122970215, + "instruction_rms": 0.0034676678115988712, + "raw_apical_rms": 0.09340542900169638, + "teaching_rms": 0.0034676678122970215, + "traffic_rms": 0.09333919046443108 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.898492633742535e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5208968764574846e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018313658323218743, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18212127259254454 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09888428449630737, + "mean_feedback_forward_cosine": 0.9980354251400116, + "mean_feedback_forward_relative_error": 0.06152387756493784, + "min_feedback_forward_cosine": 0.9951443672180176 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034759140903340606, + "instruction_rms": 0.0034759140902410157, + "raw_apical_rms": 0.09347412529018984, + "teaching_rms": 0.0034759140903340606, + "traffic_rms": 0.09340736925357711 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.2133837752089676e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.52450409723183e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018215467396750798, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.18039082668304443 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0951661467552185, + "mean_feedback_forward_cosine": 0.9981779360002087, + "mean_feedback_forward_relative_error": 0.05924614542914975, + "min_feedback_forward_cosine": 0.9955019950866699 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003423960539554479, + "instruction_rms": 0.0034239605396800634, + "raw_apical_rms": 0.09385243812729338, + "teaching_rms": 0.003423960539554479, + "traffic_rms": 0.09378770250891492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.9836078375165016e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5208947390851188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01958067184831516, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1735227191925049 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09132730215787888, + "mean_feedback_forward_cosine": 0.9983135788671432, + "mean_feedback_forward_relative_error": 0.056999864897901015, + "min_feedback_forward_cosine": 0.9958555102348328 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003445312271472041, + "instruction_rms": 0.0034453122716488374, + "raw_apical_rms": 0.0939710927944859, + "teaching_rms": 0.003445312271472041, + "traffic_rms": 0.09390568598544434 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.2083194646002084e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522992588818411e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020160802750244577, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.17735280702590941 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0876811072230339, + "mean_feedback_forward_cosine": 0.9984409097702273, + "mean_feedback_forward_relative_error": 0.05480683036148548, + "min_feedback_forward_cosine": 0.99617999792099 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035048535103925867, + "instruction_rms": 0.003504853510751594, + "raw_apical_rms": 0.09413456511708417, + "teaching_rms": 0.0035048535103925867, + "traffic_rms": 0.09406693658399976 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.163624550164059e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522649357981428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01950703249191764, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17650808246612548 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08402995020151138, + "mean_feedback_forward_cosine": 0.9985610842704773, + "mean_feedback_forward_relative_error": 0.052653420716524124, + "min_feedback_forward_cosine": 0.9964897036552429 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035416739590402417, + "instruction_rms": 0.0035416739589272674, + "raw_apical_rms": 0.09457344166876108, + "teaching_rms": 0.0035416739590402417, + "traffic_rms": 0.0945047379531391 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 5.586554863157289e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5200544741022107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01950934226676978, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17624187963962554 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08052390068769455, + "mean_feedback_forward_cosine": 0.998673592844317, + "mean_feedback_forward_relative_error": 0.05055269454756091, + "min_feedback_forward_cosine": 0.9967756271362305 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003540208157372695, + "instruction_rms": 0.0035402081569593695, + "raw_apical_rms": 0.09483157929021507, + "teaching_rms": 0.003540208157372695, + "traffic_rms": 0.09476320547439462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.452772100511538e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.518896175044665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01939992738971707, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.18187842219352723 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07720277458429337, + "mean_feedback_forward_cosine": 0.9987735517563359, + "mean_feedback_forward_relative_error": 0.04861524544896618, + "min_feedback_forward_cosine": 0.9970353841781616 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035003158300454316, + "instruction_rms": 0.0035003158309198012, + "raw_apical_rms": 0.09498571333500043, + "teaching_rms": 0.0035003158300454316, + "traffic_rms": 0.09491866178358407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.529064057123833e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5288657977678663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018480898283810714, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.1762250036621094 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07416512817144394, + "mean_feedback_forward_cosine": 0.9988657274553853, + "mean_feedback_forward_relative_error": 0.046751628900247234, + "min_feedback_forward_cosine": 0.9972637891769409 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034554718629002915, + "instruction_rms": 0.0034554718634361562, + "raw_apical_rms": 0.09519143930321672, + "teaching_rms": 0.0034554718629002915, + "traffic_rms": 0.09512648723722626 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.018802440304171e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5163553027358353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018844094197898863, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.17268958324909212 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07120780646800995, + "mean_feedback_forward_cosine": 0.9989516985031867, + "mean_feedback_forward_relative_error": 0.044948042670805606, + "min_feedback_forward_cosine": 0.9974775314331055 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003470594085441904, + "instruction_rms": 0.003470594086108436, + "raw_apical_rms": 0.09537300016442023, + "teaching_rms": 0.003470594085441904, + "traffic_rms": 0.0953077111123998 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.989027502233512e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5199447587948385e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01951974860847415, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.17510864750385285 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06835093349218369, + "mean_feedback_forward_cosine": 0.9990294710282357, + "mean_feedback_forward_relative_error": 0.04324998676536545, + "min_feedback_forward_cosine": 0.9976750016212463 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003442444643512692, + "instruction_rms": 0.0034424446438425355, + "raw_apical_rms": 0.09562175925960778, + "teaching_rms": 0.003442444643512692, + "traffic_rms": 0.09555776369150733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.6041826351483905e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.520097017286646e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019163110445721183, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.17469935647964477 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06559786200523376, + "mean_feedback_forward_cosine": 0.9991019137444035, + "mean_feedback_forward_relative_error": 0.041603886612480684, + "min_feedback_forward_cosine": 0.9978584051132202 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035098417467668364, + "instruction_rms": 0.003509841746491489, + "raw_apical_rms": 0.09607821031605751, + "teaching_rms": 0.0035098417467668364, + "traffic_rms": 0.09601196820992938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.5210232829949746e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5167766973070668e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01847280692125062, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1742268820810318 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06288089603185654, + "mean_feedback_forward_cosine": 0.9991696150072159, + "mean_feedback_forward_relative_error": 0.04000675068387101, + "min_feedback_forward_cosine": 0.9980316162109375 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034711323112744876, + "instruction_rms": 0.003471132312907595, + "raw_apical_rms": 0.09580615766435122, + "teaching_rms": 0.0034711323112744876, + "traffic_rms": 0.09574090896005594 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.436546774400085e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5224432948121803e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01807439206409788, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.16974857542037963 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.060315243899822235, + "mean_feedback_forward_cosine": 0.9992318499472833, + "mean_feedback_forward_relative_error": 0.03848021101927565, + "min_feedback_forward_cosine": 0.9981886148452759 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034625979422622188, + "instruction_rms": 0.003462597942201401, + "raw_apical_rms": 0.09608253547263627, + "teaching_rms": 0.0034625979422622188, + "traffic_rms": 0.09601792823747263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.3853660258719174e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5194168004595284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018175094498645634, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1708846864056587 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05790635570883751, + "mean_feedback_forward_cosine": 0.9992905066859338, + "mean_feedback_forward_relative_error": 0.036987309374155536, + "min_feedback_forward_cosine": 0.9983300566673279 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034659804125877205, + "instruction_rms": 0.0034659804125936338, + "raw_apical_rms": 0.09654511939410211, + "teaching_rms": 0.0034659804125877205, + "traffic_rms": 0.09648072978963708 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.4871489756606024e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514022440685767e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01872765565633019, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.17357911764621733 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05558721348643303, + "mean_feedback_forward_cosine": 0.9993443796711583, + "mean_feedback_forward_relative_error": 0.03555818403800649, + "min_feedback_forward_cosine": 0.9984607696533203 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034345960635000262, + "instruction_rms": 0.003434596063429914, + "raw_apical_rms": 0.09618863003047651, + "teaching_rms": 0.0034345960635000262, + "traffic_rms": 0.09612500039861203 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 3.0324098787559706e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5221265707216064e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018433742315666427, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.16764934017658234 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.053250689059495926, + "mean_feedback_forward_cosine": 0.9993948205824821, + "mean_feedback_forward_relative_error": 0.03416652184340262, + "min_feedback_forward_cosine": 0.9985877871513367 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034707142104435703, + "instruction_rms": 0.003470714209793029, + "raw_apical_rms": 0.09665036808699193, + "teaching_rms": 0.0034707142104435703, + "traffic_rms": 0.09658558029907076 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.3690496675499162e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5178760290844713e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018240615313675095, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.17118308273792268 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051056429743766785, + "mean_feedback_forward_cosine": 0.999440279699141, + "mean_feedback_forward_relative_error": 0.032861349653572805, + "min_feedback_forward_cosine": 0.9987015724182129 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034431129607801108, + "instruction_rms": 0.0034431129605066806, + "raw_apical_rms": 0.09721646057055212, + "teaching_rms": 0.0034431129607801108, + "traffic_rms": 0.09715313312332086 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 2.8187693956738258e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5190232767124194e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018749776611060146, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1709734722328186 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.049032460898160934, + "mean_feedback_forward_cosine": 0.9994814280540713, + "mean_feedback_forward_relative_error": 0.03162932227696142, + "min_feedback_forward_cosine": 0.9988020658493042 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003426611747302319, + "instruction_rms": 0.0034266117469159176, + "raw_apical_rms": 0.09758027006267173, + "teaching_rms": 0.003426611747302319, + "traffic_rms": 0.09751818909311871 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.413447385331892e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5193136086737388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018263243377247003, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.16499940060138701 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04703202471137047, + "mean_feedback_forward_cosine": 0.9995214516116727, + "mean_feedback_forward_relative_error": 0.030385937421552597, + "min_feedback_forward_cosine": 0.9988974332809448 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003425232360922412, + "instruction_rms": 0.0034252323605393886, + "raw_apical_rms": 0.09769198648610412, + "teaching_rms": 0.003425232360922412, + "traffic_rms": 0.09762998887938472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.9548216196626527e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5194927176303573e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017329593991227972, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16814869140625 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04509647563099861, + "mean_feedback_forward_cosine": 0.9995579546497714, + "mean_feedback_forward_relative_error": 0.029204544250763232, + "min_feedback_forward_cosine": 0.9989863634109497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003442762255018781, + "instruction_rms": 0.003442762255456184, + "raw_apical_rms": 0.09748693259128102, + "teaching_rms": 0.003442762255018781, + "traffic_rms": 0.09742396515429015 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.1700883574312684e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5252748217119797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01751308314702741, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1678281867980957 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.043238986283540726, + "mean_feedback_forward_cosine": 0.9995921158021496, + "mean_feedback_forward_relative_error": 0.028056915458892624, + "min_feedback_forward_cosine": 0.9990680813789368 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034615357179880258, + "instruction_rms": 0.003461535718166754, + "raw_apical_rms": 0.09756635091870378, + "teaching_rms": 0.0034615357179880258, + "traffic_rms": 0.09750295308892072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.310136247909213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5115054103527494e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01689349675678171, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1657409055352211 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.041482746601104736, + "mean_feedback_forward_cosine": 0.9996230063899871, + "mean_feedback_forward_relative_error": 0.026976262399506186, + "min_feedback_forward_cosine": 0.9991425275802612 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003444376064539832, + "instruction_rms": 0.003444376063942611, + "raw_apical_rms": 0.09796741979656859, + "teaching_rms": 0.003444376064539832, + "traffic_rms": 0.09790479833583576 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 5.119829538102749e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514643716629963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01734774110521595, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1643344660282135 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03971974179148674, + "mean_feedback_forward_cosine": 0.999652624130249, + "mean_feedback_forward_relative_error": 0.025895704335022356, + "min_feedback_forward_cosine": 0.9992138147354126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003515184189588737, + "instruction_rms": 0.0035151841893170045, + "raw_apical_rms": 0.0977453934640979, + "teaching_rms": 0.003515184189588737, + "traffic_rms": 0.0976801818866119 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.8098964811060796e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512423987414053e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01750532272126192, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1694715497303009 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03811245039105415, + "mean_feedback_forward_cosine": 0.9996787021237035, + "mean_feedback_forward_relative_error": 0.024905373823017843, + "min_feedback_forward_cosine": 0.999275803565979 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034469692828211, + "instruction_rms": 0.0034469692834442783, + "raw_apical_rms": 0.09823427550377863, + "teaching_rms": 0.0034469692828211, + "traffic_rms": 0.09817163928363351 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.857382407408295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5208523550200853e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01681604897758859, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1653107776927948 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0365268774330616, + "mean_feedback_forward_cosine": 0.9997029227595176, + "mean_feedback_forward_relative_error": 0.023948671596665538, + "min_feedback_forward_cosine": 0.9993346929550171 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034284203106171238, + "instruction_rms": 0.0034284203107027428, + "raw_apical_rms": 0.09867610619125185, + "teaching_rms": 0.0034284203106171238, + "traffic_rms": 0.09861440303033983 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.7250487139317556e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5077889404986472e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017965978167755036, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16640377036094667 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03507483750581741, + "mean_feedback_forward_cosine": 0.9997243823543671, + "mean_feedback_forward_relative_error": 0.023067279358304316, + "min_feedback_forward_cosine": 0.9993864893913269 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033423867646511724, + "instruction_rms": 0.0033423867648472503, + "raw_apical_rms": 0.09913136384998657, + "teaching_rms": 0.0033423867646511724, + "traffic_rms": 0.09907280568496654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.5637388979002935e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.516675588861783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017970925337275706, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16079451736450195 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03370649740099907, + "mean_feedback_forward_cosine": 0.9997442345465383, + "mean_feedback_forward_relative_error": 0.02222104301495898, + "min_feedback_forward_cosine": 0.9994333386421204 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003327080507195754, + "instruction_rms": 0.003327080507794785, + "raw_apical_rms": 0.09896504524560934, + "teaching_rms": 0.003327080507195754, + "traffic_rms": 0.09890729176005182 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.8892172088699226e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5076107550516124e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017027639971131398, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.16067632440805435 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03369491547346115, + "mean_feedback_forward_cosine": 0.9997444114377422, + "mean_feedback_forward_relative_error": 0.022211969079029177, + "min_feedback_forward_cosine": 0.999433696269989 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025155797792103646, + "instruction_rms": 0.002515579779132268, + "raw_apical_rms": 0.09902059953675202, + "teaching_rms": 0.0025155797792103646, + "traffic_rms": 0.09898664992752854 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.3049994619414065e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.504262383047355e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017186265549443417, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.08979547102928162 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03368911147117615, + "mean_feedback_forward_cosine": 0.999744507574266, + "mean_feedback_forward_relative_error": 0.022206049036955642, + "min_feedback_forward_cosine": 0.9994338750839233 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002234980681785057, + "instruction_rms": 0.0022349806810016202, + "raw_apical_rms": 0.098945154892924, + "teaching_rms": 0.002234980681785057, + "traffic_rms": 0.09891820477415945 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.323275248907521e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5069525601307564e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016625602146777663, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06746735813200473 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03368388116359711, + "mean_feedback_forward_cosine": 0.9997445825607546, + "mean_feedback_forward_relative_error": 0.02220067246666839, + "min_feedback_forward_cosine": 0.9994341135025024 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002070015851135544, + "instruction_rms": 0.002070015850555554, + "raw_apical_rms": 0.09889842510153583, + "teaching_rms": 0.002070015851135544, + "traffic_rms": 0.09887461728716397 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.299914380112568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5071429828098774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016410034565511114, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05679080193191767 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033678654581308365, + "mean_feedback_forward_cosine": 0.9997446902336613, + "mean_feedback_forward_relative_error": 0.02219541614214259, + "min_feedback_forward_cosine": 0.9994341135025024 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020812366229444137, + "instruction_rms": 0.0020812366228650987, + "raw_apical_rms": 0.09883578838904135, + "teaching_rms": 0.0020812366229444137, + "traffic_rms": 0.09881181220868573 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.6960227706449587e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507026781695462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01698622182792127, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.0515896087551117 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033673521131277084, + "mean_feedback_forward_cosine": 0.9997447729110718, + "mean_feedback_forward_relative_error": 0.02219033325391431, + "min_feedback_forward_cosine": 0.9994344711303711 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00203953780505874, + "instruction_rms": 0.002039537804548174, + "raw_apical_rms": 0.0988177521742726, + "teaching_rms": 0.00203953780505874, + "traffic_rms": 0.09879456827947249 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 5.700387362139736e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5068242328728935e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01669186849696991, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04738149812757969 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033668894320726395, + "mean_feedback_forward_cosine": 0.9997448652021347, + "mean_feedback_forward_relative_error": 0.022185512382777467, + "min_feedback_forward_cosine": 0.9994345903396606 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001955486812312254, + "instruction_rms": 0.001955486812009368, + "raw_apical_rms": 0.09873248369964849, + "teaching_rms": 0.001955486812312254, + "traffic_rms": 0.09871114437139716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.6895365812815726e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5104291520894208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016793788370509835, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04181470906853676 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033663950860500336, + "mean_feedback_forward_cosine": 0.9997449324977014, + "mean_feedback_forward_relative_error": 0.02218072828386099, + "min_feedback_forward_cosine": 0.9994348287582397 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002016420973616901, + "instruction_rms": 0.0020164209740549514, + "raw_apical_rms": 0.09873626751206291, + "teaching_rms": 0.002016420973616901, + "traffic_rms": 0.09871335214502651 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.4757872075251726e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.508164372748245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01690313283835775, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.04257393829703331 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03365926444530487, + "mean_feedback_forward_cosine": 0.9997450247887643, + "mean_feedback_forward_relative_error": 0.022175970847808545, + "min_feedback_forward_cosine": 0.9994349479675293 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001960354561209463, + "instruction_rms": 0.001960354560757627, + "raw_apical_rms": 0.09869638569040602, + "teaching_rms": 0.001960354561209463, + "traffic_rms": 0.09867469293684333 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.6254459817873794e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5047314543704612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016562018205474514, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.038318599182367324 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03365436568856239, + "mean_feedback_forward_cosine": 0.999745117079827, + "mean_feedback_forward_relative_error": 0.02217107650733763, + "min_feedback_forward_cosine": 0.9994350671768188 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002000307671734041, + "instruction_rms": 0.0020003076722440653, + "raw_apical_rms": 0.09870695599531946, + "teaching_rms": 0.002000307671734041, + "traffic_rms": 0.09868398039036168 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.2460023880485096e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5049035263034137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016861056559934185, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.0373265215241909 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03364964947104454, + "mean_feedback_forward_cosine": 0.9997451997572376, + "mean_feedback_forward_relative_error": 0.022166369332661553, + "min_feedback_forward_cosine": 0.999435305595398 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019329316008195178, + "instruction_rms": 0.001932931600735467, + "raw_apical_rms": 0.09867720001343648, + "teaching_rms": 0.0019329316008195178, + "traffic_rms": 0.09865574711492302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.918306828471941e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5048085820017315e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017061846977221226, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03438850516736507 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03364493325352669, + "mean_feedback_forward_cosine": 0.9997452728209957, + "mean_feedback_forward_relative_error": 0.02216171715108137, + "min_feedback_forward_cosine": 0.9994354248046875 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019093082474223619, + "instruction_rms": 0.001909308247367793, + "raw_apical_rms": 0.09860141010577825, + "teaching_rms": 0.0019093082474223619, + "traffic_rms": 0.09858013825512603 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.6214292837021276e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5077294557431272e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01656580144197753, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.0330116683447361 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033640310168266296, + "mean_feedback_forward_cosine": 0.9997453708802501, + "mean_feedback_forward_relative_error": 0.022157171786191962, + "min_feedback_forward_cosine": 0.9994356036186218 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001911112559351246, + "instruction_rms": 0.001911112559347709, + "raw_apical_rms": 0.09856708346859766, + "teaching_rms": 0.001911112559351246, + "traffic_rms": 0.09854569224042958 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.27534140815078e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5108777237808912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016891072717017595, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.03186257752418518 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033635251224040985, + "mean_feedback_forward_cosine": 0.9997454650940434, + "mean_feedback_forward_relative_error": 0.022152457296127272, + "min_feedback_forward_cosine": 0.9994356632232666 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018997861360157335, + "instruction_rms": 0.001899786135733846, + "raw_apical_rms": 0.09853573937144966, + "teaching_rms": 0.0018997861360157335, + "traffic_rms": 0.09851423397726856 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.630928884807872e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5090837019823764e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017390287515859002, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.030026931683421135 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03363063558936119, + "mean_feedback_forward_cosine": 0.9997455650760282, + "mean_feedback_forward_relative_error": 0.02214795453173499, + "min_feedback_forward_cosine": 0.9994359016418457 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001824119984538829, + "instruction_rms": 0.001824119984637395, + "raw_apical_rms": 0.09852454909000531, + "teaching_rms": 0.001824119984538829, + "traffic_rms": 0.09850484174041621 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.376932594141181e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.511786134028625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016590847596317994, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02798965295046568 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033626068383455276, + "mean_feedback_forward_cosine": 0.9997456362170558, + "mean_feedback_forward_relative_error": 0.022143485565339367, + "min_feedback_forward_cosine": 0.9994360208511353 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001837547651316045, + "instruction_rms": 0.0018375476514276119, + "raw_apical_rms": 0.09848647112076932, + "teaching_rms": 0.001837547651316045, + "traffic_rms": 0.09846597995239743 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.021713889163948e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5129353813668213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01697469995628945, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.027223200811743738 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03362118825316429, + "mean_feedback_forward_cosine": 0.9997457188944663, + "mean_feedback_forward_relative_error": 0.022138860437177842, + "min_feedback_forward_cosine": 0.9994361400604248 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018624103074087073, + "instruction_rms": 0.0018624103070506224, + "raw_apical_rms": 0.09847486169514422, + "teaching_rms": 0.0018624103074087073, + "traffic_rms": 0.09845416110232068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.3602349225911894e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5133615924960857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01670648100639187, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.026649290096759797 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03361663967370987, + "mean_feedback_forward_cosine": 0.9997458054173377, + "mean_feedback_forward_relative_error": 0.02213443877295621, + "min_feedback_forward_cosine": 0.9994363784790039 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018177894473187955, + "instruction_rms": 0.0018177894475962899, + "raw_apical_rms": 0.09843954420182029, + "teaching_rms": 0.0018177894473187955, + "traffic_rms": 0.09841946931701281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.189184548726189e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5112037043679986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01652430601417022, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.024802653382122515 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03361181542277336, + "mean_feedback_forward_cosine": 0.9997458823265568, + "mean_feedback_forward_relative_error": 0.022129854412689325, + "min_feedback_forward_cosine": 0.9994363784790039 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001846084160118297, + "instruction_rms": 0.0018460841597463572, + "raw_apical_rms": 0.0984229459518161, + "teaching_rms": 0.001846084160118297, + "traffic_rms": 0.09840259313026754 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.0474415758117175e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5109034247487467e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017178483696749437, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.0250774456769228 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03360702097415924, + "mean_feedback_forward_cosine": 0.9997459919221939, + "mean_feedback_forward_relative_error": 0.02212545263671106, + "min_feedback_forward_cosine": 0.9994367361068726 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017378667343276428, + "instruction_rms": 0.0017378667344857901, + "raw_apical_rms": 0.09839323336558864, + "teaching_rms": 0.0017378667343276428, + "traffic_rms": 0.09837453043963171 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.507539586029452e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5098816895733553e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016789284838536228, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.022657332441806795 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033602274954319, + "mean_feedback_forward_cosine": 0.9997460745996044, + "mean_feedback_forward_relative_error": 0.022120956421619462, + "min_feedback_forward_cosine": 0.9994368553161621 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017874540886220066, + "instruction_rms": 0.0017874540890151294, + "raw_apical_rms": 0.09832726511314333, + "teaching_rms": 0.0017874540886220066, + "traffic_rms": 0.09830738431591088 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.144525161426827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5065207920935208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017052075946436353, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.022775477136969566 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03359761834144592, + "mean_feedback_forward_cosine": 0.9997461418951711, + "mean_feedback_forward_relative_error": 0.02211661014946238, + "min_feedback_forward_cosine": 0.9994369745254517 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001685905954877595, + "instruction_rms": 0.0016859059548532044, + "raw_apical_rms": 0.09829528258820518, + "teaching_rms": 0.001685905954877595, + "traffic_rms": 0.09827727892954943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.148975290580893e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5077906976214916e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01687566242869036, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.020612371822595597 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03359293192625046, + "mean_feedback_forward_cosine": 0.999746230340773, + "mean_feedback_forward_relative_error": 0.022112091883055626, + "min_feedback_forward_cosine": 0.999437153339386 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017261941708247553, + "instruction_rms": 0.0017261941712981856, + "raw_apical_rms": 0.09825673203904728, + "teaching_rms": 0.0017261941708247553, + "traffic_rms": 0.09823718786461569 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.188960017221689e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5107916533187797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016920253913493877, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.021344696981608867 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03358827158808708, + "mean_feedback_forward_cosine": 0.9997463168636445, + "mean_feedback_forward_relative_error": 0.022107676783156012, + "min_feedback_forward_cosine": 0.9994372129440308 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001694066485788303, + "instruction_rms": 0.0016940664858105432, + "raw_apical_rms": 0.0982178426291435, + "teaching_rms": 0.001694066485788303, + "traffic_rms": 0.09819914397612879 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.482267002003937e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5035406403036503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017060581974225438, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.020410545910894872 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03358400613069534, + "mean_feedback_forward_cosine": 0.9997463956955941, + "mean_feedback_forward_relative_error": 0.022103574742833452, + "min_feedback_forward_cosine": 0.9994373917579651 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016310455171500357, + "instruction_rms": 0.0016310455174278914, + "raw_apical_rms": 0.09817094437035466, + "teaching_rms": 0.0016310455171500357, + "traffic_rms": 0.09815399189915751 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.64203004924002e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.498872109185774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016858379729896508, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.018607610389888285 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03357977420091629, + "mean_feedback_forward_cosine": 0.9997464918321178, + "mean_feedback_forward_relative_error": 0.022099388462881887, + "min_feedback_forward_cosine": 0.9994375705718994 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016288643216268957, + "instruction_rms": 0.001628864321668804, + "raw_apical_rms": 0.098136995211163, + "teaching_rms": 0.0016288643216268957, + "traffic_rms": 0.09811988544116251 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.921443874650322e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.497979198978486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017132574005115395, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.01782615624934435 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033574849367141724, + "mean_feedback_forward_cosine": 0.9997465725867979, + "mean_feedback_forward_relative_error": 0.022094827280530045, + "min_feedback_forward_cosine": 0.9994378089904785 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016856051769386096, + "instruction_rms": 0.001685605176678328, + "raw_apical_rms": 0.09808741994623163, + "teaching_rms": 0.0016856051769386096, + "traffic_rms": 0.09806890919491446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.4941423854870663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512230283690769e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017247794900485065, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.01872813436359167 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03357018157839775, + "mean_feedback_forward_cosine": 0.9997466860278961, + "mean_feedback_forward_relative_error": 0.022090584400200074, + "min_feedback_forward_cosine": 0.9994380474090576 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001613412239006114, + "instruction_rms": 0.0016134122389234356, + "raw_apical_rms": 0.09809247999997249, + "teaching_rms": 0.001613412239006114, + "traffic_rms": 0.09807515246959388 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.966822331198273e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.511276753485453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017143375548244386, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.017819994999170304 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03356596827507019, + "mean_feedback_forward_cosine": 0.9997467437098103, + "mean_feedback_forward_relative_error": 0.022086514370335687, + "min_feedback_forward_cosine": 0.9994380474090576 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001564545914479394, + "instruction_rms": 0.0015645459147263585, + "raw_apical_rms": 0.09810397155459925, + "teaching_rms": 0.001564545914479394, + "traffic_rms": 0.09808790570229337 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.583021251391983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.516690890979968e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017041482295694566, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.016292895752489565 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033561427146196365, + "mean_feedback_forward_cosine": 0.9997468090826466, + "mean_feedback_forward_relative_error": 0.02208245791975529, + "min_feedback_forward_cosine": 0.9994382858276367 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015503156777161044, + "instruction_rms": 0.0015503156773543436, + "raw_apical_rms": 0.09808399036718388, + "teaching_rms": 0.0015503156777161044, + "traffic_rms": 0.09806829689235912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.15442053483639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507983267816016e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017249684926626378, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.01625606567889452 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03355669975280762, + "mean_feedback_forward_cosine": 0.999746895605518, + "mean_feedback_forward_relative_error": 0.02207810063696196, + "min_feedback_forward_cosine": 0.9994383454322815 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001589629891166118, + "instruction_rms": 0.001589629890417969, + "raw_apical_rms": 0.09805560151738811, + "teaching_rms": 0.001589629891166118, + "traffic_rms": 0.09803876374243303 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.4578921887250544e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.515333139225125e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01708194413848096, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.016294221063926816 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03355200216174126, + "mean_feedback_forward_cosine": 0.9997469936647723, + "mean_feedback_forward_relative_error": 0.022073830072317394, + "min_feedback_forward_cosine": 0.9994385242462158 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015597470082573273, + "instruction_rms": 0.0015597470084517642, + "raw_apical_rms": 0.09806876543254142, + "teaching_rms": 0.0015597470082573273, + "traffic_rms": 0.09805228231546412 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 5.11607218811605e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5126213060985655e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01791325396731845, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.015560092131197453 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033547110855579376, + "mean_feedback_forward_cosine": 0.9997470898012961, + "mean_feedback_forward_relative_error": 0.022069225523380503, + "min_feedback_forward_cosine": 0.9994386434555054 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016545587865269284, + "instruction_rms": 0.001654558787143388, + "raw_apical_rms": 0.09804469062816437, + "teaching_rms": 0.0016545587865269284, + "traffic_rms": 0.09802649184942097 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.999416264832689e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509364299692674e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017231682173616647, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.01624549331843853 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03354316204786301, + "mean_feedback_forward_cosine": 0.9997471551741323, + "mean_feedback_forward_relative_error": 0.022065332399741296, + "min_feedback_forward_cosine": 0.9994388818740845 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001419893063309232, + "instruction_rms": 0.0014198930635276723, + "raw_apical_rms": 0.09801270399905807, + "teaching_rms": 0.001419893063309232, + "traffic_rms": 0.09799854957409332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.564810041074452e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5081127227142167e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016762355791938605, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013408799584805965 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033538758754730225, + "mean_feedback_forward_cosine": 0.9997472455424647, + "mean_feedback_forward_relative_error": 0.02206124270695352, + "min_feedback_forward_cosine": 0.999439001083374 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014698420868666459, + "instruction_rms": 0.001469842086642212, + "raw_apical_rms": 0.0980235655117867, + "teaching_rms": 0.0014698420868666459, + "traffic_rms": 0.09800869389546062 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.1951484064384204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5113732514480768e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017163864761790697, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.01408297025412321 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0335342139005661, + "mean_feedback_forward_cosine": 0.9997473262971447, + "mean_feedback_forward_relative_error": 0.022056753101247932, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00154979082290428, + "instruction_rms": 0.0015497908224634336, + "raw_apical_rms": 0.09804923163422168, + "teaching_rms": 0.00154979082290428, + "traffic_rms": 0.09803230841865349 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.6809685017024094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5126607201581415e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016779777062986206, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.014305016270875931 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033529285341501236, + "mean_feedback_forward_cosine": 0.9997474089745553, + "mean_feedback_forward_relative_error": 0.022052083885477435, + "min_feedback_forward_cosine": 0.9994392395019531 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015508211834305277, + "instruction_rms": 0.0015508211834875168, + "raw_apical_rms": 0.0980319862680842, + "teaching_rms": 0.0015508211834305277, + "traffic_rms": 0.09801552868363742 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.4990706271619274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5123010981834098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01711876487188448, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.014552696642279626 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352479636669159, + "mean_feedback_forward_cosine": 0.9997475108792705, + "mean_feedback_forward_relative_error": 0.022047825998836947, + "min_feedback_forward_cosine": 0.9994393587112427 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014629426341457184, + "instruction_rms": 0.0014629426340389048, + "raw_apical_rms": 0.09806354465040687, + "teaching_rms": 0.0014629426341457184, + "traffic_rms": 0.09804828476743588 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.277175236211406e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5069503454750045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017516926863231738, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.013265480405688286 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03352031111717224, + "mean_feedback_forward_cosine": 0.9997475954794115, + "mean_feedback_forward_relative_error": 0.022043652742380095, + "min_feedback_forward_cosine": 0.9994396567344666 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001444701276559143, + "instruction_rms": 0.0014447012764091474, + "raw_apical_rms": 0.09809343087247323, + "teaching_rms": 0.001444701276559143, + "traffic_rms": 0.09807907097560523 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.554451749072541e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5060700977711647e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017030207414552495, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.01280905030965805 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03351566195487976, + "mean_feedback_forward_cosine": 0.9997476608522476, + "mean_feedback_forward_relative_error": 0.022039273017716986, + "min_feedback_forward_cosine": 0.9994397759437561 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014640756723155559, + "instruction_rms": 0.001464075672303349, + "raw_apical_rms": 0.09806848510301143, + "teaching_rms": 0.0014640756723155559, + "traffic_rms": 0.09805333247914437 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.802994541501903e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5117463996103862e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017154134597832094, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.013464438293129205 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03351116552948952, + "mean_feedback_forward_cosine": 0.9997477608342324, + "mean_feedback_forward_relative_error": 0.022035094849284616, + "min_feedback_forward_cosine": 0.9994399547576904 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014012798931144752, + "instruction_rms": 0.0014012798928951813, + "raw_apical_rms": 0.0980432424253836, + "teaching_rms": 0.0014012798931144752, + "traffic_rms": 0.09802864912695801 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.604452996659347e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509178630625995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01690935256791718, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012629328387826681 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033506475389003754, + "mean_feedback_forward_cosine": 0.9997478627389477, + "mean_feedback_forward_relative_error": 0.02203046844430989, + "min_feedback_forward_cosine": 0.99944007396698 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015045285009746398, + "instruction_rms": 0.0015045285011594262, + "raw_apical_rms": 0.09796912910751018, + "teaching_rms": 0.0015045285009746398, + "traffic_rms": 0.09795328745695527 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.216471438667213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.515502519434141e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017074015476496108, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.012571220416724682 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033501289784908295, + "mean_feedback_forward_cosine": 0.9997479492618192, + "mean_feedback_forward_relative_error": 0.0220256901886915, + "min_feedback_forward_cosine": 0.9994403123855591 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001522503870667616, + "instruction_rms": 0.0015225038707435941, + "raw_apical_rms": 0.09800897737591947, + "teaching_rms": 0.001522503870667616, + "traffic_rms": 0.09799264943358482 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.5757289512948773e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514528674171317e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01724100692401263, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.013056165408939124 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03349626809358597, + "mean_feedback_forward_cosine": 0.9997480684711088, + "mean_feedback_forward_relative_error": 0.02202099674351273, + "min_feedback_forward_cosine": 0.9994404315948486 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001476427784006252, + "instruction_rms": 0.0014764277843166034, + "raw_apical_rms": 0.09807307073383509, + "teaching_rms": 0.001476427784006252, + "traffic_rms": 0.09805755934533104 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 3.722982810927533e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5010595778049722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01731819006236269, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.012149872734434903 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03349178284406662, + "mean_feedback_forward_cosine": 0.9997481492257887, + "mean_feedback_forward_relative_error": 0.02201673972810949, + "min_feedback_forward_cosine": 0.9994405508041382 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013871223344801762, + "instruction_rms": 0.0013871223348780803, + "raw_apical_rms": 0.0980632985115755, + "teaching_rms": 0.0013871223344801762, + "traffic_rms": 0.09804934546084002 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.3981934538805945e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5038121894655725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017159862157832453, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.011330395812988281 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03348664566874504, + "mean_feedback_forward_cosine": 0.9997482434395821, + "mean_feedback_forward_relative_error": 0.02201197502173243, + "min_feedback_forward_cosine": 0.9994406700134277 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001443142921656888, + "instruction_rms": 0.001443142921663592, + "raw_apical_rms": 0.09798071573341327, + "teaching_rms": 0.001443142921656888, + "traffic_rms": 0.09796510416943062 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.854843200201685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.508584427134651e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016935549227715965, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.012074666630923749 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03348210081458092, + "mean_feedback_forward_cosine": 0.9997483357306449, + "mean_feedback_forward_relative_error": 0.02200756493895765, + "min_feedback_forward_cosine": 0.9994409084320068 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013895581288645442, + "instruction_rms": 0.0013895581291020627, + "raw_apical_rms": 0.09793144550711869, + "teaching_rms": 0.0013895581288645442, + "traffic_rms": 0.09791734044888825 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.682980758730082e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512139101513327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016884070136261326, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.01120645552277565 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03347731754183769, + "mean_feedback_forward_cosine": 0.999748412639864, + "mean_feedback_forward_relative_error": 0.02200296955303319, + "min_feedback_forward_cosine": 0.9994410276412964 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014036982092907647, + "instruction_rms": 0.001403698209259251, + "raw_apical_rms": 0.0979298131640237, + "teaching_rms": 0.0014036982092907647, + "traffic_rms": 0.09791559665381201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.694131223141994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5137285936989785e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01685449019195481, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.011195772307813167 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033472105860710144, + "mean_feedback_forward_cosine": 0.9997485183900402, + "mean_feedback_forward_relative_error": 0.021997922190254735, + "min_feedback_forward_cosine": 0.9994412064552307 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014533156741845864, + "instruction_rms": 0.0014533156742920056, + "raw_apical_rms": 0.0978506531843855, + "teaching_rms": 0.0014533156741845864, + "traffic_rms": 0.09783481506422335 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 5.081500376036274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5097093962619986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01696975916787381, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.011413695975542069 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346719220280647, + "mean_feedback_forward_cosine": 0.9997486260629469, + "mean_feedback_forward_relative_error": 0.021993313089854295, + "min_feedback_forward_cosine": 0.9994413256645203 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013752153971263897, + "instruction_rms": 0.001375215397130992, + "raw_apical_rms": 0.09784760791050572, + "teaching_rms": 0.0013752153971263897, + "traffic_rms": 0.0978335315983768 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.319190611250633e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5142346341287133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016812764126789302, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.011152344236299395 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346230834722519, + "mean_feedback_forward_cosine": 0.9997487145085489, + "mean_feedback_forward_relative_error": 0.021988661494106054, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013855820995497471, + "instruction_rms": 0.0013855820997256017, + "raw_apical_rms": 0.09783614863319473, + "teaching_rms": 0.0013855820995497471, + "traffic_rms": 0.09782185740494999 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.284744647975458e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5040538969479908e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01739325477733304, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.010847439026683569 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346225246787071, + "mean_feedback_forward_cosine": 0.9997487125858184, + "mean_feedback_forward_relative_error": 0.021988567475590012, + "min_feedback_forward_cosine": 0.999441385269165 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012627994642478047, + "instruction_rms": 0.001262799464439568, + "raw_apical_rms": 0.09781154208800588, + "teaching_rms": 0.0012627994642478047, + "traffic_rms": 0.09779910673922346 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.1713615672733795e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5053889575708106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016798549941149907, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.009891995935589076 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346221148967743, + "mean_feedback_forward_cosine": 0.9997487183540098, + "mean_feedback_forward_relative_error": 0.021988483296046333, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011656046977669501, + "instruction_rms": 0.0011656046977164707, + "raw_apical_rms": 0.0978120013255861, + "teaching_rms": 0.0011656046977669501, + "traffic_rms": 0.09780150295329486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.3496164892076195e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.50399198497686e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01696662331090375, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008626569434180855 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346218168735504, + "mean_feedback_forward_cosine": 0.9997487241222013, + "mean_feedback_forward_relative_error": 0.021988402961963607, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011890742599111371, + "instruction_rms": 0.0011890742597635497, + "raw_apical_rms": 0.0978232141037857, + "teaching_rms": 0.0011890742599111371, + "traffic_rms": 0.09781184142863504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.086465210662027e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5059610380657376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016791044374417884, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0090260252353549 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346214443445206, + "mean_feedback_forward_cosine": 0.9997487221994708, + "mean_feedback_forward_relative_error": 0.021988322177240924, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011345420937767118, + "instruction_rms": 0.0011345420937356558, + "raw_apical_rms": 0.09782639311149878, + "teaching_rms": 0.0011345420937767118, + "traffic_rms": 0.09781619437712097 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 3.439311768715925e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5061918505059912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01694866419538979, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.008467365986406803 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346210718154907, + "mean_feedback_forward_cosine": 0.9997487395040451, + "mean_feedback_forward_relative_error": 0.02198824226375549, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001113794391875477, + "instruction_rms": 0.0011137943921179826, + "raw_apical_rms": 0.09782851301810036, + "teaching_rms": 0.001113794391875477, + "traffic_rms": 0.09781891132357805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.212223814192839e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5104748915189617e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01691682249410719, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.008258016648441554 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033462073653936386, + "mean_feedback_forward_cosine": 0.9997487279676622, + "mean_feedback_forward_relative_error": 0.021988164843811144, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010856588273521616, + "instruction_rms": 0.001085658827630451, + "raw_apical_rms": 0.09783515710505178, + "teaching_rms": 0.0010856588273521616, + "traffic_rms": 0.09782574070922827 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.828781996479847e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5080737064066813e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01692112620992937, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.008091233827769757 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0334620364010334, + "mean_feedback_forward_cosine": 0.9997487491176974, + "mean_feedback_forward_relative_error": 0.021988088054762732, + "min_feedback_forward_cosine": 0.9994414448738098 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010948563759436987, + "instruction_rms": 0.0010948563755085203, + "raw_apical_rms": 0.09783567864079627, + "teaching_rms": 0.0010948563759436987, + "traffic_rms": 0.09782592052474515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.423794663921399e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5055341123606702e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017262641666157236, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.00838008071988821 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346199169754982, + "mean_feedback_forward_cosine": 0.9997487241222013, + "mean_feedback_forward_relative_error": 0.021988007074762738, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010771409061681974, + "instruction_rms": 0.0010771409062401744, + "raw_apical_rms": 0.0978332363415038, + "teaching_rms": 0.0010771409061681974, + "traffic_rms": 0.09782434486383458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.008446552461464e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5099393147021254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01745256597373089, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.007727677876055241 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461958169937134, + "mean_feedback_forward_cosine": 0.9997487318131232, + "mean_feedback_forward_relative_error": 0.02198792339092301, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001086709963225053, + "instruction_rms": 0.0010867099633507106, + "raw_apical_rms": 0.09783867886201776, + "teaching_rms": 0.001086709963225053, + "traffic_rms": 0.09782940347644901 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.396456176891661e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5076985132706176e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016784451798148838, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007944510354101658 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461928367614746, + "mean_feedback_forward_cosine": 0.9997487414267755, + "mean_feedback_forward_relative_error": 0.021987847608303832, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010714697546107615, + "instruction_rms": 0.0010714697549424747, + "raw_apical_rms": 0.09782980667919249, + "teaching_rms": 0.0010714697546107615, + "traffic_rms": 0.09782091029741573 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.35951896804454e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5115644588727587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016918933062033407, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007814029241502285 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346189111471176, + "mean_feedback_forward_cosine": 0.9997487202767403, + "mean_feedback_forward_relative_error": 0.021987770714106097, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010750854510190757, + "instruction_rms": 0.001075085450921433, + "raw_apical_rms": 0.09783121627296004, + "teaching_rms": 0.0010750854510190757, + "traffic_rms": 0.09782185692501347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 4.134049902404513e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5107020310150084e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017315641057948656, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007849760715514421 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346184641122818, + "mean_feedback_forward_cosine": 0.9997487452722364, + "mean_feedback_forward_relative_error": 0.02198768980921276, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010565535435045933, + "instruction_rms": 0.0010565535438089389, + "raw_apical_rms": 0.09783664921366254, + "teaching_rms": 0.0010565535435045933, + "traffic_rms": 0.09782716789189061 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 5.1760615349394357e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5095533003680496e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017002576357360837, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007620867423005402 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0334618054330349, + "mean_feedback_forward_cosine": 0.9997487414267755, + "mean_feedback_forward_relative_error": 0.02198760935495938, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011361077138248039, + "instruction_rms": 0.0011361077139719535, + "raw_apical_rms": 0.09783838464370832, + "teaching_rms": 0.0011361077138248039, + "traffic_rms": 0.09782749476453297 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.1231515979234246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5161534972212377e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016842319704360814, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.008356548052579165 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346177563071251, + "mean_feedback_forward_cosine": 0.9997487337358536, + "mean_feedback_forward_relative_error": 0.02198753442855612, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001049725086730029, + "instruction_rms": 0.0010497250866412632, + "raw_apical_rms": 0.09783915388104597, + "teaching_rms": 0.001049725086730029, + "traffic_rms": 0.09783015708963229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.993336861185526e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5101557478936923e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016974089176749328, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.00754912966042757 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461734652519226, + "mean_feedback_forward_cosine": 0.9997487452722364, + "mean_feedback_forward_relative_error": 0.021987456633078474, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010017802899781078, + "instruction_rms": 0.0010017802900814932, + "raw_apical_rms": 0.09783456225986217, + "teaching_rms": 0.0010017802899781078, + "traffic_rms": 0.09782609117104504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.302124168020782e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5066358247465608e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017284933498124648, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.007420491152405739 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346170112490654, + "mean_feedback_forward_cosine": 0.9997487395040451, + "mean_feedback_forward_relative_error": 0.021987383989917654, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010754016924618794, + "instruction_rms": 0.0010754016923535837, + "raw_apical_rms": 0.0978322069064871, + "teaching_rms": 0.0010754016924618794, + "traffic_rms": 0.09782284616085915 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.2061483185307225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5064743445908164e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016881738907112896, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.007952698883712292 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346167877316475, + "mean_feedback_forward_cosine": 0.9997487356585841, + "mean_feedback_forward_relative_error": 0.021987308597853108, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010462393238791549, + "instruction_rms": 0.0010462393236629233, + "raw_apical_rms": 0.09783076811060853, + "teaching_rms": 0.0010462393238791549, + "traffic_rms": 0.09782127132477632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.573439937871626e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510049105652229e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01673006555052853, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007624274515211582 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461641520261765, + "mean_feedback_forward_cosine": 0.999748747194967, + "mean_feedback_forward_relative_error": 0.021987233160724564, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010599048693950488, + "instruction_rms": 0.0010599048694836025, + "raw_apical_rms": 0.09783048645458725, + "teaching_rms": 0.0010599048693950488, + "traffic_rms": 0.09782095755943312 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.530748671209949e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512513481918644e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01679067914162356, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.007430742657184601 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346160426735878, + "mean_feedback_forward_cosine": 0.9997487337358536, + "mean_feedback_forward_relative_error": 0.021987151519785964, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001090258096307509, + "instruction_rms": 0.0010902580965397907, + "raw_apical_rms": 0.09783287102212372, + "teaching_rms": 0.001090258096307509, + "traffic_rms": 0.09782283470814632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.448238755117018e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5082142166720544e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016973422327446433, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.007761554230898618 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461567014455795, + "mean_feedback_forward_cosine": 0.9997487356585841, + "mean_feedback_forward_relative_error": 0.021987073258647034, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010361441117856117, + "instruction_rms": 0.0010361441119491346, + "raw_apical_rms": 0.09783507326714401, + "teaching_rms": 0.0010361441117856117, + "traffic_rms": 0.09782607048663394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.603339220921043e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.504418026834398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017220707179922825, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0073820266579836605 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461540937423706, + "mean_feedback_forward_cosine": 0.9997487298903927, + "mean_feedback_forward_relative_error": 0.021987001787150098, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010581639306103314, + "instruction_rms": 0.0010581639307537122, + "raw_apical_rms": 0.0978356464520185, + "teaching_rms": 0.0010581639306103314, + "traffic_rms": 0.09782638473144684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 3.638755430301899e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5047260247517532e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017062984804974875, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.007644726791009307 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346150740981102, + "mean_feedback_forward_cosine": 0.9997487318131232, + "mean_feedback_forward_relative_error": 0.021986927762026747, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009760029309151166, + "instruction_rms": 0.000976002930900774, + "raw_apical_rms": 0.09783484360127037, + "teaching_rms": 0.0009760029309151166, + "traffic_rms": 0.0978270697815345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.8522394163464924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5103518522548882e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016780035153168664, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.007235477397739888 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346147760748863, + "mean_feedback_forward_cosine": 0.999748743349506, + "mean_feedback_forward_relative_error": 0.02198685480341796, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009662784440275465, + "instruction_rms": 0.0009662784439448497, + "raw_apical_rms": 0.09783197907038528, + "teaching_rms": 0.0009662784440275465, + "traffic_rms": 0.09782392931172172 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.241037032670647e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5014342513461312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01692315325943027, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006871422180980444 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461444079875946, + "mean_feedback_forward_cosine": 0.9997487529631583, + "mean_feedback_forward_relative_error": 0.021986780102334676, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009719774731417203, + "instruction_rms": 0.0009719774730486973, + "raw_apical_rms": 0.0978241366307454, + "teaching_rms": 0.0009719774731417203, + "traffic_rms": 0.09781621591915375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.0986513738516805e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5088732906988475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01700439719410832, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006632893297970295 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346141055226326, + "mean_feedback_forward_cosine": 0.9997487529631583, + "mean_feedback_forward_relative_error": 0.021986707173768553, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009915397081059483, + "instruction_rms": 0.0009915397078150132, + "raw_apical_rms": 0.09782724883088141, + "teaching_rms": 0.0009915397081059483, + "traffic_rms": 0.09781909791471065 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.625195077461285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5044420949467117e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016958505549520653, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006991697119027376 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461373299360275, + "mean_feedback_forward_cosine": 0.9997487491176974, + "mean_feedback_forward_relative_error": 0.02198663998335119, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000995704109985665, + "instruction_rms": 0.0009957041100551534, + "raw_apical_rms": 0.09782923817024504, + "teaching_rms": 0.000995704109985665, + "traffic_rms": 0.09782111449033665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.355988991733284e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5085686613012365e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016645759200534443, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.007189493736326695 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461350947618484, + "mean_feedback_forward_cosine": 0.9997487568086193, + "mean_feedback_forward_relative_error": 0.02198656699469974, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001006525418502741, + "instruction_rms": 0.001006525418333147, + "raw_apical_rms": 0.09782644878028326, + "teaching_rms": 0.001006525418502741, + "traffic_rms": 0.09781757090149104 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.5491670874753254e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.506210570383535e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016682542837920943, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.007161466164216399 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0334613211452961, + "mean_feedback_forward_cosine": 0.9997487202767403, + "mean_feedback_forward_relative_error": 0.021986490010374015, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010801429886031805, + "instruction_rms": 0.0010801429886137942, + "raw_apical_rms": 0.09782250730340863, + "teaching_rms": 0.0010801429886031805, + "traffic_rms": 0.09781249528764666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 4.249496754482607e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5144858074680852e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01708466029510744, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.007427461018115282 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461280167102814, + "mean_feedback_forward_cosine": 0.9997487568086193, + "mean_feedback_forward_relative_error": 0.021986417877938477, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009916715141545633, + "instruction_rms": 0.0009916715142240494, + "raw_apical_rms": 0.09782169292640143, + "teaching_rms": 0.0009916715141545633, + "traffic_rms": 0.09781363689347464 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.8052135664656817e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510617857276934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017088593902598228, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006919085892960429 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033461254090070724, + "mean_feedback_forward_cosine": 0.9997487395040451, + "mean_feedback_forward_relative_error": 0.021986340112503496, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009413612900152019, + "instruction_rms": 0.0009413612901111625, + "raw_apical_rms": 0.09781401264572645, + "teaching_rms": 0.0009413612900152019, + "traffic_rms": 0.09780646181596962 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.980387575097666e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.506946008045113e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016787501075676822, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006770315288156271 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346122428774834, + "mean_feedback_forward_cosine": 0.9997487548858889, + "mean_feedback_forward_relative_error": 0.021986261355660616, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010673207880264207, + "instruction_rms": 0.0010673207876248946, + "raw_apical_rms": 0.09782127142068035, + "teaching_rms": 0.0010673207880264207, + "traffic_rms": 0.09781169657966567 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.368048678315972e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510700502353046e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01681871970317486, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.007532179688513279 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346119448542595, + "mean_feedback_forward_cosine": 0.9997487510404279, + "mean_feedback_forward_relative_error": 0.021986187405643925, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009357309471480661, + "instruction_rms": 0.0009357309467854641, + "raw_apical_rms": 0.09782740540822249, + "teaching_rms": 0.0009357309471480661, + "traffic_rms": 0.09781986145143987 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.7833959964928e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5080013155996984e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016794049810857133, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006479265240728855 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346116095781326, + "mean_feedback_forward_cosine": 0.9997487510404279, + "mean_feedback_forward_relative_error": 0.021986113801117865, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009663652246616081, + "instruction_rms": 0.0009663652245007103, + "raw_apical_rms": 0.0978204077636458, + "teaching_rms": 0.0009663652246616081, + "traffic_rms": 0.09781255123935426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.8576442424064226e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5099077511611616e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016970622105779074, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0066555366459488865 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346112370491028, + "mean_feedback_forward_cosine": 0.9997487568086193, + "mean_feedback_forward_relative_error": 0.02198604329098617, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009652929816333927, + "instruction_rms": 0.0009652929815246807, + "raw_apical_rms": 0.0978185281057343, + "teaching_rms": 0.0009652929816333927, + "traffic_rms": 0.0978105267949347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.747338084281182e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.50454985702368e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017840878331515785, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006970881218090653 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346110135316849, + "mean_feedback_forward_cosine": 0.9997487356585841, + "mean_feedback_forward_relative_error": 0.021985968049134937, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009259488320416489, + "instruction_rms": 0.0009259488322631424, + "raw_apical_rms": 0.09780553882606946, + "teaching_rms": 0.0009259488320416489, + "traffic_rms": 0.09779806829615217 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.910393514023781e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5038509800024222e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017531173102358293, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006169848254993558 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0334610715508461, + "mean_feedback_forward_cosine": 0.9997487568086193, + "mean_feedback_forward_relative_error": 0.02198589856045381, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009279392139499063, + "instruction_rms": 0.0009279392138865069, + "raw_apical_rms": 0.09780321761663453, + "teaching_rms": 0.0009279392139499063, + "traffic_rms": 0.09779613334362237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.70856270137665e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5090362275841283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017778360770421085, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006411543487235904 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346104174852371, + "mean_feedback_forward_cosine": 0.9997487644995412, + "mean_feedback_forward_relative_error": 0.021985832571742997, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009244338779973663, + "instruction_rms": 0.0009244338778394026, + "raw_apical_rms": 0.09780525854433622, + "teaching_rms": 0.0009244338779973663, + "traffic_rms": 0.09779857417875858 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.078692583675547e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507371671336707e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016831820046871154, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006312604299113155 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346100449562073, + "mean_feedback_forward_cosine": 0.9997487529631583, + "mean_feedback_forward_relative_error": 0.021985756008014563, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009952868612912353, + "instruction_rms": 0.0009952868617366098, + "raw_apical_rms": 0.09781164904073858, + "teaching_rms": 0.0009952868612912353, + "traffic_rms": 0.09780282498690947 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.495152967047943e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5051920508166232e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016862775833676034, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006927484558969736 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346097469329834, + "mean_feedback_forward_cosine": 0.9997487625768108, + "mean_feedback_forward_relative_error": 0.021985680706078006, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010001924115647428, + "instruction_rms": 0.001000192411457422, + "raw_apical_rms": 0.09781407837757926, + "teaching_rms": 0.0010001924115647428, + "traffic_rms": 0.0978056220346848 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954791843891144, + "maximum_absolute_post_projection_soma_slope": 3.9040084942598696e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5119222619748722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017259255880668404, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006934199496656656 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346094489097595, + "mean_feedback_forward_cosine": 0.9997487529631583, + "mean_feedback_forward_relative_error": 0.021985611938420805, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009321066423410907, + "instruction_rms": 0.0009321066419995321, + "raw_apical_rms": 0.09780209840272498, + "teaching_rms": 0.0009321066423410907, + "traffic_rms": 0.09779479776646227 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.7572085886949935e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5055680626628026e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016960205859536837, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.006575519463196397 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346092253923416, + "mean_feedback_forward_cosine": 0.9997487683450023, + "mean_feedback_forward_relative_error": 0.02198554413212884, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000900506226772395, + "instruction_rms": 0.0009005062271540262, + "raw_apical_rms": 0.09779733017239384, + "teaching_rms": 0.000900506226772395, + "traffic_rms": 0.0977905018086833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.08888460867729e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507185101074406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01689790304693667, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.005963441434800625 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346088528633118, + "mean_feedback_forward_cosine": 0.9997487702677327, + "mean_feedback_forward_relative_error": 0.02198547533442897, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009851622218666665, + "instruction_rms": 0.0009851622216119583, + "raw_apical_rms": 0.09779682578700669, + "teaching_rms": 0.0009851622218666665, + "traffic_rms": 0.09778810048604404 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.944888860019091e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5065916673994602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017377358288078407, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006992345408499241 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033460866659879684, + "mean_feedback_forward_cosine": 0.999748747194967, + "mean_feedback_forward_relative_error": 0.021985403907996033, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009290499116851508, + "instruction_rms": 0.0009290499117290523, + "raw_apical_rms": 0.09779464307563751, + "teaching_rms": 0.0009290499116851508, + "traffic_rms": 0.09778714761893346 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.749213289301224e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.508527756759286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017598486230607797, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006231327361315489 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0334608368575573, + "mean_feedback_forward_cosine": 0.9997487510404279, + "mean_feedback_forward_relative_error": 0.02198533506523217, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009073109658764565, + "instruction_rms": 0.0009073109660308787, + "raw_apical_rms": 0.09779970175351536, + "teaching_rms": 0.0009073109658764565, + "traffic_rms": 0.09779240395957206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 3.745383381215106e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.511437394037558e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016980700202892083, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.006254670864790678 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346080332994461, + "mean_feedback_forward_cosine": 0.9997487625768108, + "mean_feedback_forward_relative_error": 0.021985268400561427, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009166585566266625, + "instruction_rms": 0.0009166585566469095, + "raw_apical_rms": 0.09780111630093874, + "teaching_rms": 0.0009166585566266625, + "traffic_rms": 0.09779356376358765 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.413101706290945e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5093607668419292e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017270588708930487, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0064557369216531515 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346078097820282, + "mean_feedback_forward_cosine": 0.9997487818041155, + "mean_feedback_forward_relative_error": 0.021985196643659183, + "min_feedback_forward_cosine": 0.9994415044784546 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009294155787038035, + "instruction_rms": 0.0009294155789720674, + "raw_apical_rms": 0.09780693064540853, + "teaching_rms": 0.0009294155787038035, + "traffic_rms": 0.0977994375154307 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.55607747085196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512677255169995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016577441297848897, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.006207835003584623 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033460747450590134, + "mean_feedback_forward_cosine": 0.9997487721904632, + "mean_feedback_forward_relative_error": 0.021985124361010328, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009688627092430869, + "instruction_rms": 0.0009688627092061012, + "raw_apical_rms": 0.09779826413321763, + "teaching_rms": 0.0009688627092430869, + "traffic_rms": 0.09778942790853662 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.1416541307626176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5076610708138134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016909866614855468, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.006491875571906567 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346073254942894, + "mean_feedback_forward_cosine": 0.9997487721904632, + "mean_feedback_forward_relative_error": 0.021985051537593527, + "min_feedback_forward_cosine": 0.9994415640830994 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008416567995618732, + "instruction_rms": 0.0008416567994213161, + "raw_apical_rms": 0.09779882987086966, + "teaching_rms": 0.0008416567995618732, + "traffic_rms": 0.09779270400300039 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954792588949203, + "maximum_absolute_post_projection_soma_slope": 4.4801652165915584e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5048195173204572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01696822285612318, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.005883447072654963 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03346070647239685, + "mean_feedback_forward_cosine": 0.9997487548858889, + "mean_feedback_forward_relative_error": 0.02198498111758982, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009350290640017631, + "instruction_rms": 0.0009350290640471836, + "raw_apical_rms": 0.09779380957240542, + "teaching_rms": 0.0009350290640017631, + "traffic_rms": 0.09778609100371703 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954794079065323, + "maximum_absolute_post_projection_soma_slope": 4.0128121270299744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5086613180592683e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016864896405583543, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0062462591500207785 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033460669219493866, + "mean_feedback_forward_cosine": 0.9997487606540802, + "mean_feedback_forward_relative_error": 0.021984909991583517, + "min_feedback_forward_cosine": 0.9994416236877441 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009550125014581149, + "instruction_rms": 0.0009550125015064038, + "raw_apical_rms": 0.09778722979516817, + "teaching_rms": 0.0009550125014581149, + "traffic_rms": 0.0977791756914359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.11954793334007263, + "maximum_absolute_post_projection_soma_slope": 5.132462632673196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5166592499155975e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017351259657444322, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.00648252774566412 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.921, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.39785928955078126, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2529336320, + "peak_memory_reserved_bytes": 3821010944, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 6.043068623284853e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 6.043068623284853e-08, + "min_residual_soma_slope": -4.4811869770455814e-08, + "mse": 5.701875412029035e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.211621889319629e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.211621889319629e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1383540630340576, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.043192386627197266, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 14827.987923622131, + "train_wall_s": 14824.164831876755 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.024588478729128838, + 0.014591400511562824, + 0.010299528017640114, + 0.010247637517750263, + 0.005952616687864065, + 0.006740528624504805, + 0.004027113318443298, + 0.005008075851947069, + 0.0031860945746302605, + 0.004283188842236996, + 0.002583871828392148, + 0.005423903465270996, + 0.003592056455090642, + 0.0044009508565068245, + 0.0029432540759444237, + 0.003715700702741742, + 0.0025049929972738028, + 0.0034612512681633234, + 0.0021376078948378563, + 0.002984041813760996, + 0.001826574676670134, + 0.004144418518990278, + 0.0029650835786014795, + 0.0037945127114653587, + 0.0025767760816961527, + 0.0033097954001277685, + 0.0023035630583763123, + 0.0029016095213592052, + 0.002129862317815423, + 0.0027866438031196594, + 0.002112960210070014 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.1323377639055252, + 0.07711678743362427, + 0.03510509431362152, + 0.05386357754468918, + 0.016026047989726067, + 0.03617874160408974, + 0.009136411361396313, + 0.026053285226225853, + 0.006380654405802488, + 0.023591075092554092, + 0.004578462336212397, + 0.029599502682685852, + 0.00799194909632206, + 0.022512374445796013, + 0.005925713572651148, + 0.01995171047747135, + 0.004546777345240116, + 0.018205545842647552, + 0.003493580035865307, + 0.016529854387044907, + 0.002779953181743622, + 0.02215592935681343, + 0.005763883702456951, + 0.020242298021912575, + 0.004539282526820898, + 0.017432086169719696, + 0.0037844351027160883, + 0.015237879008054733, + 0.0032716202549636364, + 0.014591528102755547, + 0.003002380020916462 + ], + "unscaled_traffic_rms": [ + 0.7432036995887756, + 0.7568469047546387, + 1.1735650300979614, + 0.7610068321228027, + 1.4857354164123535, + 0.7452474236488342, + 1.7631051540374756, + 0.7688974142074585, + 1.9973465204238892, + 0.7262388467788696, + 2.2574145793914795, + 0.732972264289856, + 1.797837495803833, + 0.78196120262146, + 1.9867677688598633, + 0.7449387907981873, + 2.2037525177001953, + 0.7604828476905823, + 2.447469711303711, + 0.7220975756645203, + 2.628209352493286, + 0.7482274174690247, + 2.0576984882354736, + 0.7498185634613037, + 2.270646095275879, + 0.7594720125198364, + 2.4347760677337646, + 0.7616832852363586, + 2.604045867919922, + 0.7639073133468628, + 2.815047025680542 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684206618894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 72745580756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 2745665106100224 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json new file mode 100644 index 0000000..a3687e2 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json @@ -0,0 +1,6861 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5013, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9996227920055389, + "feedback_forward_cosine": [ + 0.9998266696929932, + 0.9998146295547485, + 0.9997915029525757, + 0.9997797012329102, + 0.9997129440307617, + 0.9996341466903687, + 0.9997661709785461, + 0.9996988773345947, + 0.9997620582580566, + 0.9996570944786072, + 0.9997298717498779, + 0.9998093247413635, + 0.9997180104255676, + 0.9996817708015442, + 0.9997682571411133, + 0.9997138977050781, + 0.9997902512550354, + 0.9997091293334961, + 0.9997783899307251, + 0.9996909499168396, + 0.9997871518135071, + 0.9998667240142822, + 0.9998387098312378, + 0.9998230934143066, + 0.9998358488082886, + 0.9997856616973877, + 0.9998124837875366, + 0.9997082948684692, + 0.9997628927230835, + 0.9995109438896179, + 0.999983549118042 + ], + "feedback_forward_norm_ratio": [ + 1.0003786087036133, + 0.9999864101409912, + 0.9992262125015259, + 1.00118088722229, + 0.998736560344696, + 1.002174735069275, + 0.9988030791282654, + 1.0015902519226074, + 0.9985690116882324, + 1.0013126134872437, + 0.9991462230682373, + 1.0006859302520752, + 0.9994093775749207, + 1.0011804103851318, + 0.9994674324989319, + 1.0011271238327026, + 0.9996803998947144, + 1.001161813735962, + 0.9996950626373291, + 1.001160740852356, + 0.9995452761650085, + 1.000220775604248, + 0.9999011158943176, + 1.0008418560028076, + 0.9997518062591553, + 1.0011337995529175, + 0.9997069835662842, + 1.0013108253479004, + 0.9997100234031677, + 1.0013885498046875, + 1.0022026300430298 + ], + "feedback_forward_relative_error": [ + 0.018618928268551826, + 0.019253486767411232, + 0.020427050068974495, + 0.021034903824329376, + 0.023977361619472504, + 0.027164874598383904, + 0.021647583693265915, + 0.024609200656414032, + 0.02184344455599785, + 0.02623707242310047, + 0.02325104922056198, + 0.019542844966053963, + 0.023751458153128624, + 0.02527329884469509, + 0.02152971178293228, + 0.023958662524819374, + 0.020484240725636482, + 0.02416081354022026, + 0.02105093002319336, + 0.024906929582357407, + 0.020637262612581253, + 0.016326529905200005, + 0.017961647361516953, + 0.018835319206118584, + 0.0181178729981184, + 0.02074573002755642, + 0.019360698759555817, + 0.02420783042907715, + 0.021775471046566963, + 0.03132582828402519, + 0.006159801501780748 + ], + "innovation_negative_gradient_cosine": [ + 0.9996398091316223, + 0.9996058940887451, + 0.99964439868927, + 0.999598503112793, + 0.9996539354324341, + 0.9995397925376892, + 0.9996646046638489, + 0.9996017217636108, + 0.9996776580810547, + 0.9996016025543213, + 0.999685525894165, + 0.9996574521064758, + 0.999701738357544, + 0.9995663166046143, + 0.9997093677520752, + 0.9996194839477539, + 0.9997212886810303, + 0.9996551871299744, + 0.9997247457504272, + 0.9996870756149292, + 0.9997378587722778, + 0.9997754693031311, + 0.9997869729995728, + 0.9997658133506775, + 0.9998039603233337, + 0.999817967414856, + 0.9998483657836914, + 0.9998794794082642, + 0.9998995065689087, + 0.999940037727356, + 0.7504194974899292 + ], + "instruction_negative_gradient_cosine": [ + 0.9996398091316223, + 0.9996058940887451, + 0.99964439868927, + 0.999598503112793, + 0.9996539354324341, + 0.9995397925376892, + 0.9996645450592041, + 0.9996017217636108, + 0.9996776580810547, + 0.9996016025543213, + 0.999685525894165, + 0.999657392501831, + 0.9997017979621887, + 0.9995663166046143, + 0.9997093677520752, + 0.9996194243431091, + 0.9997212886810303, + 0.9996551871299744, + 0.9997246861457825, + 0.999687135219574, + 0.9997378587722778, + 0.9997754693031311, + 0.9997869729995728, + 0.9997657537460327, + 0.9998039603233337, + 0.9998179078102112, + 0.9998483657836914, + 0.9998794794082642, + 0.9998995065689087, + 0.999940037727356, + 0.7685295343399048 + ], + "matched_negative_gradient_cosine": [ + 0.018930844962596893, + 0.02241251990199089, + 0.011943921446800232, + 0.03673360496759415, + 0.006199301220476627, + -0.02070508897304535, + 0.004725276492536068, + -0.041707202792167664, + 0.03994016349315643, + 0.03537318855524063, + 0.048001304268836975, + 0.06243788078427315, + 0.03955027833580971, + 0.06811034679412842, + 0.03466589003801346, + 0.06164909526705742, + 0.046681422740221024, + 0.12694469094276428, + 0.02356584370136261, + 0.10797655582427979, + 0.01714702695608139, + 0.03305324539542198, + -0.052473850548267365, + 0.10837972164154053, + 0.005332642700523138, + 0.16741801798343658, + 0.017147552222013474, + 0.16580763459205627, + 0.036554574966430664, + 0.04411827772855759, + 0.09979928284883499 + ], + "max_norm_match_direction_error": 0.002277970314025879, + "max_norm_match_relative_error": 1.498547561595842e-07, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.021554123805535418, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.07146559655666351, + "max_absolute_post_projection_soma_slope": 1.0606905576082681e-08, + "max_absolute_pre_projection_soma_slope": 0.07146559655666351, + "max_positive_post_projection_soma_slope": 1.0606905576082681e-08, + "min_post_projection_soma_slope": -8.543406870842318e-09, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.467356403610835e-08, + "pre_projection_traffic_rms_ratio": 0.017169242892281625 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.01716924289226287, + "raw_negative_gradient_cosine": [ + 0.018930844962596893, + 0.02241251990199089, + 0.011943922378122807, + 0.03673360496759415, + 0.006199301220476627, + -0.0207050908356905, + 0.004725276958197355, + -0.04170720651745796, + 0.03994016349315643, + 0.03537318855524063, + 0.04800130054354668, + 0.06243788078427315, + 0.03955027833580971, + 0.06811034679412842, + 0.03466589003801346, + 0.06164909526705742, + 0.046681422740221024, + 0.12694469094276428, + 0.02356584556400776, + 0.10797655582427979, + 0.01714702881872654, + 0.03305324539542198, + -0.05247385799884796, + 0.10837972164154053, + 0.005332643631845713, + 0.16741801798343658, + 0.017147554084658623, + 0.16580763459205627, + 0.036554574966430664, + 0.044118281453847885, + 0.09979875385761261 + ], + "teaching_negative_gradient_cosine": [ + 0.9996398091316223, + 0.9996058940887451, + 0.99964439868927, + 0.999598503112793, + 0.9996539354324341, + 0.9995397925376892, + 0.9996646046638489, + 0.9996017217636108, + 0.9996776580810547, + 0.9996016025543213, + 0.999685525894165, + 0.9996574521064758, + 0.999701738357544, + 0.9995663166046143, + 0.9997093677520752, + 0.9996194839477539, + 0.9997212886810303, + 0.9996551871299744, + 0.9997247457504272, + 0.9996870756149292, + 0.9997378587722778, + 0.9997754693031311, + 0.9997869729995728, + 0.9997658133506775, + 0.9998039603233337, + 0.999817967414856, + 0.9998483657836914, + 0.9998794794082642, + 0.9998995065689087, + 0.999940037727356, + 0.7504194974899292 + ], + "traffic_instruction_rms_ratio": 44.26736325044913, + "used_negative_gradient_cosine": [ + 0.9996398091316223, + 0.9996058940887451, + 0.99964439868927, + 0.999598503112793, + 0.9996539354324341, + 0.9995397925376892, + 0.9996646046638489, + 0.9996017217636108, + 0.9996776580810547, + 0.9996016025543213, + 0.999685525894165, + 0.9996574521064758, + 0.999701738357544, + 0.9995663166046143, + 0.9997093677520752, + 0.9996194839477539, + 0.9997212886810303, + 0.9996551871299744, + 0.9997247457504272, + 0.9996870756149292, + 0.9997378587722778, + 0.9997754693031311, + 0.9997869729995728, + 0.9997658133506775, + 0.9998039603233337, + 0.999817967414856, + 0.9998483657836914, + 0.9998794794082642, + 0.9998995065689087, + 0.999940037727356, + 0.7504194974899292 + ], + "wall_s": 0.23176050186157227 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.359028935432434, + "mean_feedback_forward_cosine": 0.21638116048228356, + "mean_feedback_forward_relative_error": 1.2513171953539695, + "min_feedback_forward_cosine": 0.08171359449625015 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0015689440774127226, + "instruction_rms": 0.001568944077005395, + "raw_apical_rms": 0.027384441832174197, + "teaching_rms": 0.0015689440774127226, + "traffic_rms": 0.027309113920130117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.166573930319828e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.023138328154589e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013457047512726689, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.0188516019058227 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.327566146850586, + "mean_feedback_forward_cosine": 0.31701424237220516, + "mean_feedback_forward_relative_error": 1.1690104257675908, + "min_feedback_forward_cosine": 0.12498265504837036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0023151978247357317, + "instruction_rms": 0.0023151978249872354, + "raw_apical_rms": 0.03159317279282856, + "teaching_rms": 0.0023151978247357317, + "traffic_rms": 0.03150716712562438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0714656189084053, + "maximum_absolute_post_projection_soma_slope": 3.529411785052616e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.836178078147449e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013728502577306671, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4014113458633424 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2820147275924683, + "mean_feedback_forward_cosine": 0.4102951413200748, + "mean_feedback_forward_relative_error": 1.087201907750099, + "min_feedback_forward_cosine": 0.1869596391916275 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031249409712210702, + "instruction_rms": 0.003124940971169802, + "raw_apical_rms": 0.037262446519259876, + "teaching_rms": 0.0031249409712210702, + "traffic_rms": 0.037130303814453064 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.5341944482070176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.752664916148555e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015238662020276414, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.0887655848884583 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.236717700958252, + "mean_feedback_forward_cosine": 0.4784102209152714, + "mean_feedback_forward_relative_error": 1.0225722559036747, + "min_feedback_forward_cosine": 0.2463923990726471 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003368224748347706, + "instruction_rms": 0.0033682247480265395, + "raw_apical_rms": 0.04022405698360309, + "teaching_rms": 0.003368224748347706, + "traffic_rms": 0.04008171859118678 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.797032732677508e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7262246808453166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01405065507174546, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.8817265232276916 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1967111825942993, + "mean_feedback_forward_cosine": 0.5311479856891017, + "mean_feedback_forward_relative_error": 0.9698550451186395, + "min_feedback_forward_cosine": 0.30145370960235596 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033948441496080973, + "instruction_rms": 0.0033948441491979033, + "raw_apical_rms": 0.0418392732684939, + "teaching_rms": 0.0033948441496080973, + "traffic_rms": 0.041700224559935176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.26482663126626e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7032469829814485e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015228620355479124, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.7550507375526428 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1632848978042603, + "mean_feedback_forward_cosine": 0.5752259637078931, + "mean_feedback_forward_relative_error": 0.9231321782835068, + "min_feedback_forward_cosine": 0.34760457277297974 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003479446511401995, + "instruction_rms": 0.003479446513266296, + "raw_apical_rms": 0.04310546706742094, + "teaching_rms": 0.003479446511401995, + "traffic_rms": 0.04296349270536215 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.992172009110618e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6970980854575346e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01552344969720946, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.6805254464530944 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1323214769363403, + "mean_feedback_forward_cosine": 0.6145065042280382, + "mean_feedback_forward_relative_error": 0.8791621029376984, + "min_feedback_forward_cosine": 0.3874923586845398 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035217726659655317, + "instruction_rms": 0.0035217726660450766, + "raw_apical_rms": 0.04403249099993706, + "teaching_rms": 0.0035217726659655317, + "traffic_rms": 0.04389020462136081 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.99808044448946e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.686406539880861e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017570972230950884, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6278603303432465 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.103968620300293, + "mean_feedback_forward_cosine": 0.648109265873509, + "mean_feedback_forward_relative_error": 0.839671830977163, + "min_feedback_forward_cosine": 0.4198325276374817 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035703464131065328, + "instruction_rms": 0.00357034641325869, + "raw_apical_rms": 0.04510810332447291, + "teaching_rms": 0.0035703464131065328, + "traffic_rms": 0.044965083341892545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.006622589258768e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6675527453982265e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01608658976536159, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5761027395820618 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.074819564819336, + "mean_feedback_forward_cosine": 0.6776645520041066, + "mean_feedback_forward_relative_error": 0.803543588807506, + "min_feedback_forward_cosine": 0.45070701837539673 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003525060820714708, + "instruction_rms": 0.003525060821088069, + "raw_apical_rms": 0.045850809822561636, + "teaching_rms": 0.003525060820714708, + "traffic_rms": 0.04571381135685931 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 4.192946079228932e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.654928768429143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017757043925315817, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.5437979357337952 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0463757514953613, + "mean_feedback_forward_cosine": 0.7039340215344583, + "mean_feedback_forward_relative_error": 0.7700270058647278, + "min_feedback_forward_cosine": 0.48043709993362427 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003468220437532979, + "instruction_rms": 0.00346822043729152, + "raw_apical_rms": 0.04646558855553005, + "teaching_rms": 0.003468220437532979, + "traffic_rms": 0.04633434519209877 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.013248672800728e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.661593374398277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018891471597838407, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5084304156112671 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0177077054977417, + "mean_feedback_forward_cosine": 0.7282911250668187, + "mean_feedback_forward_relative_error": 0.737410927972486, + "min_feedback_forward_cosine": 0.5099976062774658 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003481418197683415, + "instruction_rms": 0.003481418197371488, + "raw_apical_rms": 0.04694401919291556, + "teaching_rms": 0.003481418197683415, + "traffic_rms": 0.046813320010665474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.4188545328106557e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6593052960661066e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020091266190934612, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4850193161201477 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9888617396354675, + "mean_feedback_forward_cosine": 0.7506973647302196, + "mean_feedback_forward_relative_error": 0.7061494290828705, + "min_feedback_forward_cosine": 0.5377273559570312 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036126322643429826, + "instruction_rms": 0.0036126322642947594, + "raw_apical_rms": 0.0472092087437721, + "teaching_rms": 0.0036126322643429826, + "traffic_rms": 0.04706907247442703 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.2148527395747806e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6559376135540032e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02059401780472702, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4700221423149109 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9599944353103638, + "mean_feedback_forward_cosine": 0.7703979111486866, + "mean_feedback_forward_relative_error": 0.6774305085982045, + "min_feedback_forward_cosine": 0.564585268497467 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003528432517188769, + "instruction_rms": 0.00352843251735401, + "raw_apical_rms": 0.048000954012695385, + "teaching_rms": 0.003528432517188769, + "traffic_rms": 0.04786957593681927 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 3.298147888131098e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6489699692673787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020034272408130764, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.44316998596191404 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9321396946907043, + "mean_feedback_forward_cosine": 0.7884342593531455, + "mean_feedback_forward_relative_error": 0.6499206144963542, + "min_feedback_forward_cosine": 0.5900013446807861 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035830642252036495, + "instruction_rms": 0.0035830642265681782, + "raw_apical_rms": 0.048310959002821695, + "teaching_rms": 0.0035830642252036495, + "traffic_rms": 0.04817617398786338 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.238727768779427e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.661857639360773e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021262276743741006, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.42744599245071413 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9045042991638184, + "mean_feedback_forward_cosine": 0.8049306677233788, + "mean_feedback_forward_relative_error": 0.6238999366760254, + "min_feedback_forward_cosine": 0.6142401099205017 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003556955471479921, + "instruction_rms": 0.003556955470923244, + "raw_apical_rms": 0.04871852154079069, + "teaching_rms": 0.003556955471479921, + "traffic_rms": 0.04858674191570059 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.5328916681010014e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6547784543395443e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020975813899543384, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4125113959121704 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8764141201972961, + "mean_feedback_forward_cosine": 0.8199109358172263, + "mean_feedback_forward_relative_error": 0.5991476283919427, + "min_feedback_forward_cosine": 0.6378459930419922 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003551704510117151, + "instruction_rms": 0.0035517045091667445, + "raw_apical_rms": 0.04893941869978095, + "teaching_rms": 0.003551704510117151, + "traffic_rms": 0.04880854297151682 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.0586108696061274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6496897206309826e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02161130895052106, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.40041986715316774 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8499462008476257, + "mean_feedback_forward_cosine": 0.8342026606682809, + "mean_feedback_forward_relative_error": 0.5745326886253972, + "min_feedback_forward_cosine": 0.659515380859375 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003617971944839745, + "instruction_rms": 0.003617971944487428, + "raw_apical_rms": 0.0493595282590454, + "teaching_rms": 0.003617971944839745, + "traffic_rms": 0.04922483868044985 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.155868810722495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6415964649145383e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020677236035194786, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.384989180431366 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8231675624847412, + "mean_feedback_forward_cosine": 0.8468809396989884, + "mean_feedback_forward_relative_error": 0.5519563930649911, + "min_feedback_forward_cosine": 0.6805944442749023 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003553208022122026, + "instruction_rms": 0.0035532080221910857, + "raw_apical_rms": 0.04972547518575145, + "teaching_rms": 0.003553208022122026, + "traffic_rms": 0.04959630263131576 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.8159112375192308e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.644184583823938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021212845139219966, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3696881667804718 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7967016696929932, + "mean_feedback_forward_cosine": 0.8586022430850614, + "mean_feedback_forward_relative_error": 0.5301316515091927, + "min_feedback_forward_cosine": 0.7006962895393372 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003574316668101611, + "instruction_rms": 0.0035743166677613394, + "raw_apical_rms": 0.04971021291024696, + "teaching_rms": 0.003574316668101611, + "traffic_rms": 0.049579680769877336 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.699557365483997e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6434803741912808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02145337360642842, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.36057512926101687 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7712675929069519, + "mean_feedback_forward_cosine": 0.8694090151017712, + "mean_feedback_forward_relative_error": 0.5091335167807918, + "min_feedback_forward_cosine": 0.7194774746894836 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036112910793361856, + "instruction_rms": 0.0036112910792821754, + "raw_apical_rms": 0.050105810211630956, + "teaching_rms": 0.0036112910793361856, + "traffic_rms": 0.04997340475186982 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.2544470229822764e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6464941007630675e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01958559857282328, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3472319854450226 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7452573180198669, + "mean_feedback_forward_cosine": 0.8794327993546763, + "mean_feedback_forward_relative_error": 0.48886588840715345, + "min_feedback_forward_cosine": 0.7375442981719971 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036141921638518943, + "instruction_rms": 0.003614192164661703, + "raw_apical_rms": 0.05033756037191736, + "teaching_rms": 0.0036141921638518943, + "traffic_rms": 0.0502054539370443 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.255738789675888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6466105393357188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02239439965588174, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3434600825691223 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7202416658401489, + "mean_feedback_forward_cosine": 0.8885373800031601, + "mean_feedback_forward_relative_error": 0.46982263509304295, + "min_feedback_forward_cosine": 0.7548590898513794 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035751062003129765, + "instruction_rms": 0.003575106201163679, + "raw_apical_rms": 0.050367085565707785, + "teaching_rms": 0.0035751062003129765, + "traffic_rms": 0.05023780323841156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.616031480329184e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6447045558420602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020982626868283494, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.33141831305503844 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6955456137657166, + "mean_feedback_forward_cosine": 0.8971660617859133, + "mean_feedback_forward_relative_error": 0.4511003359671562, + "min_feedback_forward_cosine": 0.7713476419448853 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003589489087871081, + "instruction_rms": 0.003589489088420716, + "raw_apical_rms": 0.050725285002784296, + "teaching_rms": 0.003589489087871081, + "traffic_rms": 0.05059592366457357 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.0372857057491274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641078922667729e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02248168618368078, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.32522446617126466 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6716258525848389, + "mean_feedback_forward_cosine": 0.9048702889873136, + "mean_feedback_forward_relative_error": 0.43357097237340864, + "min_feedback_forward_cosine": 0.7864834070205688 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003599795016958122, + "instruction_rms": 0.003599795016235446, + "raw_apical_rms": 0.05090630499450526, + "teaching_rms": 0.003599795016958122, + "traffic_rms": 0.05077644478873893 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.3885496247544324e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6389962455493745e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02367017186458206, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3163902703666687 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6481041312217712, + "mean_feedback_forward_cosine": 0.912163544085718, + "mean_feedback_forward_relative_error": 0.4165429251809274, + "min_feedback_forward_cosine": 0.8010036945343018 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035793771446135185, + "instruction_rms": 0.0035793771453639977, + "raw_apical_rms": 0.05116234581715712, + "teaching_rms": 0.0035793771446135185, + "traffic_rms": 0.05103445498586097 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.0160639258601805e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.647446026643522e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021846705459387137, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31165570446014407 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6251521706581116, + "mean_feedback_forward_cosine": 0.9189614711269256, + "mean_feedback_forward_relative_error": 0.3999131860271577, + "min_feedback_forward_cosine": 0.8146384358406067 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036160961163394143, + "instruction_rms": 0.0036160961164440858, + "raw_apical_rms": 0.05122236264266877, + "teaching_rms": 0.0036160961163394143, + "traffic_rms": 0.05109193022478306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.0642112847090175e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638040829717733e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02185609805716488, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.29939852534294126 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6029499173164368, + "mean_feedback_forward_cosine": 0.9251291674952353, + "mean_feedback_forward_relative_error": 0.3841595308434579, + "min_feedback_forward_cosine": 0.8272641897201538 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035756388406752095, + "instruction_rms": 0.0035756388396603303, + "raw_apical_rms": 0.05130942585302663, + "teaching_rms": 0.0035756388406752095, + "traffic_rms": 0.05118217572795193 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.5922756325135197e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6362150913621873e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020947848142713835, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29339407252311706 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5805547833442688, + "mean_feedback_forward_cosine": 0.9307969500941615, + "mean_feedback_forward_relative_error": 0.3691556097999696, + "min_feedback_forward_cosine": 0.8396104574203491 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035719889993269527, + "instruction_rms": 0.003571988999061554, + "raw_apical_rms": 0.051782248044754946, + "teaching_rms": 0.0035719889993269527, + "traffic_rms": 0.05165644963930771 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.260048586639641e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6284546112516536e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02128249975974906, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2898948735380173 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5591834783554077, + "mean_feedback_forward_cosine": 0.9359502196311951, + "mean_feedback_forward_relative_error": 0.3550844954386834, + "min_feedback_forward_cosine": 0.8510515689849854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003492938000034598, + "instruction_rms": 0.003492937999622593, + "raw_apical_rms": 0.05199398085737323, + "teaching_rms": 0.003492938000034598, + "traffic_rms": 0.051874226666474704 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.2424932072672163e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6323404755410115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021101769614255105, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28035081473350526 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5376722812652588, + "mean_feedback_forward_cosine": 0.9408655743445119, + "mean_feedback_forward_relative_error": 0.3410439998392136, + "min_feedback_forward_cosine": 0.8620774149894714 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003564203899771304, + "instruction_rms": 0.003564203899696792, + "raw_apical_rms": 0.05252402522271201, + "teaching_rms": 0.003564203899771304, + "traffic_rms": 0.052400066261991285 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.118455976254154e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.630327828341386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020573800790122818, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2819810107803345 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5178632140159607, + "mean_feedback_forward_cosine": 0.9453231000131176, + "mean_feedback_forward_relative_error": 0.3278006286390366, + "min_feedback_forward_cosine": 0.8718243837356567 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035412261523905, + "instruction_rms": 0.003541226152104235, + "raw_apical_rms": 0.05295211851808451, + "teaching_rms": 0.0035412261523905, + "traffic_rms": 0.052831046655815644 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.871572313267734e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6210759042514302e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020325780500081613, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.271732903175354 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4983406066894531, + "mean_feedback_forward_cosine": 0.9494874073613074, + "mean_feedback_forward_relative_error": 0.31497027364469343, + "min_feedback_forward_cosine": 0.8811737298965454 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035404604517921064, + "instruction_rms": 0.003540460453805183, + "raw_apical_rms": 0.05256521727958658, + "teaching_rms": 0.0035404604517921064, + "traffic_rms": 0.05244321861362014 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.583815555430192e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.630777602858127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021537006481828675, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26568462629318235 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4798280596733093, + "mean_feedback_forward_cosine": 0.9532626386611692, + "mean_feedback_forward_relative_error": 0.3028777061450866, + "min_feedback_forward_cosine": 0.8896836042404175 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035311689701902496, + "instruction_rms": 0.0035311689708615295, + "raw_apical_rms": 0.05260948823950943, + "teaching_rms": 0.0035311689701902496, + "traffic_rms": 0.05248837250706658 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.353325522288287e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6272145647704196e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021035924458852533, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.26067846254348753 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46089527010917664, + "mean_feedback_forward_cosine": 0.9569080202810226, + "mean_feedback_forward_relative_error": 0.2907417671334359, + "min_feedback_forward_cosine": 0.8981483578681946 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035838708999212953, + "instruction_rms": 0.0035838708997651602, + "raw_apical_rms": 0.05302693666963662, + "teaching_rms": 0.0035838708999212953, + "traffic_rms": 0.052903118140250234 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.645661645317432e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.628055769819409e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019556002866465076, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25794370862960814 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.44318878650665283, + "mean_feedback_forward_cosine": 0.9601221796005003, + "mean_feedback_forward_relative_error": 0.27960770024407294, + "min_feedback_forward_cosine": 0.9056601524353027 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003492636714525397, + "instruction_rms": 0.0034926367143797866, + "raw_apical_rms": 0.053045539200379105, + "teaching_rms": 0.003492636714525397, + "traffic_rms": 0.05292757125978629 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 3.382749014235742e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6203148991470475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01954111414139787, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2521926338386536 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4259234666824341, + "mean_feedback_forward_cosine": 0.9631822589905031, + "mean_feedback_forward_relative_error": 0.26858584318430195, + "min_feedback_forward_cosine": 0.9127770662307739 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00351352174643121, + "instruction_rms": 0.0035135217473944657, + "raw_apical_rms": 0.05300544540767023, + "teaching_rms": 0.00351352174643121, + "traffic_rms": 0.05288612320398315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.3526106335557415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6233771448000343e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020034404823617482, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2552177179145813 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4095364511013031, + "mean_feedback_forward_cosine": 0.9659874285421064, + "mean_feedback_forward_relative_error": 0.2580969290387246, + "min_feedback_forward_cosine": 0.9191899299621582 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035540171829123095, + "instruction_rms": 0.0035540171831040367, + "raw_apical_rms": 0.053163167733734626, + "teaching_rms": 0.0035540171829123095, + "traffic_rms": 0.053041392546976665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.212623767012701e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.628830762196484e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020177349498023973, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2466756143426895 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3934934139251709, + "mean_feedback_forward_cosine": 0.9685390014802256, + "mean_feedback_forward_relative_error": 0.24812661616071577, + "min_feedback_forward_cosine": 0.9252965450286865 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035482394739834206, + "instruction_rms": 0.003548239473537122, + "raw_apical_rms": 0.05337888214745203, + "teaching_rms": 0.0035482394739834206, + "traffic_rms": 0.05325808728298396 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.0761033659464374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6322921325240592e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01930333037709197, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24432047214984895 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37840384244918823, + "mean_feedback_forward_cosine": 0.9709713632060636, + "mean_feedback_forward_relative_error": 0.23828263340457792, + "min_feedback_forward_cosine": 0.930831789970398 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035966518538834225, + "instruction_rms": 0.0035966518536361862, + "raw_apical_rms": 0.053553472376124144, + "teaching_rms": 0.0035966518538834225, + "traffic_rms": 0.05342942585547273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3190059650346484e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.621805526148183e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019432024424663498, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2401316197872162 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3639370799064636, + "mean_feedback_forward_cosine": 0.9731036636137194, + "mean_feedback_forward_relative_error": 0.22933144434805838, + "min_feedback_forward_cosine": 0.9359540343284607 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003464426539131027, + "instruction_rms": 0.0034644265394348753, + "raw_apical_rms": 0.05382661571582705, + "teaching_rms": 0.003464426539131027, + "traffic_rms": 0.05371212399086486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.038320883774759e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6157032582062388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019516162274940768, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23337802155971527 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34985122084617615, + "mean_feedback_forward_cosine": 0.9751064892738096, + "mean_feedback_forward_relative_error": 0.2205553547509255, + "min_feedback_forward_cosine": 0.9406843185424805 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003518671379162316, + "instruction_rms": 0.00351867137963814, + "raw_apical_rms": 0.05419758359191867, + "teaching_rms": 0.003518671379162316, + "traffic_rms": 0.054080755051409 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.807940369109474e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6208270594430585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020722597974415086, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2313947339439392 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3358475863933563, + "mean_feedback_forward_cosine": 0.9769871542530675, + "mean_feedback_forward_relative_error": 0.21202974838595237, + "min_feedback_forward_cosine": 0.9452786445617676 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003478429278219998, + "instruction_rms": 0.0034784292799613877, + "raw_apical_rms": 0.05425824730060113, + "teaching_rms": 0.003478429278219998, + "traffic_rms": 0.05414382479056838 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.628799388228799e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6072423458175056e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01980896056037481, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.23239266793727875 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32218971848487854, + "mean_feedback_forward_cosine": 0.9787662202312101, + "mean_feedback_forward_relative_error": 0.2036249695285674, + "min_feedback_forward_cosine": 0.949577808380127 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003547579124787204, + "instruction_rms": 0.003547579124956494, + "raw_apical_rms": 0.054477864704447376, + "teaching_rms": 0.003547579124787204, + "traffic_rms": 0.05435907189525733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.641603558117822e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6061506229604884e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02066779336193575, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.2314801558971405 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3090890347957611, + "mean_feedback_forward_cosine": 0.9803262102988458, + "mean_feedback_forward_relative_error": 0.1959815582921428, + "min_feedback_forward_cosine": 0.9535435438156128 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034376989958522968, + "instruction_rms": 0.0034376989942922647, + "raw_apical_rms": 0.05464901483278287, + "teaching_rms": 0.0034376989958522968, + "traffic_rms": 0.05453837628225535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0291723856757926e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6174431106958267e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020464399910602792, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22742463251113892 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2970096468925476, + "mean_feedback_forward_cosine": 0.9818116984059734, + "mean_feedback_forward_relative_error": 0.18838050985528576, + "min_feedback_forward_cosine": 0.9570397734642029 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035152441600312407, + "instruction_rms": 0.0035152441607478723, + "raw_apical_rms": 0.054818932900638564, + "teaching_rms": 0.0035152441600312407, + "traffic_rms": 0.05470334052266086 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.524999054287491e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6136291851556815e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020241933019776597, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2225275700044632 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2847195267677307, + "mean_feedback_forward_cosine": 0.9831811458833756, + "mean_feedback_forward_relative_error": 0.18112104145749922, + "min_feedback_forward_cosine": 0.9604690074920654 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034642623013427442, + "instruction_rms": 0.0034642623005248243, + "raw_apical_rms": 0.055020485280956904, + "teaching_rms": 0.0034642623013427442, + "traffic_rms": 0.05490841618240984 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 3.1338498018840255e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6085323177164533e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021792486458115424, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21858973684310912 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27325090765953064, + "mean_feedback_forward_cosine": 0.9844556393161896, + "mean_feedback_forward_relative_error": 0.17409104529407718, + "min_feedback_forward_cosine": 0.9635643362998962 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003524163215761551, + "instruction_rms": 0.003524163215956543, + "raw_apical_rms": 0.05532184571723606, + "teaching_rms": 0.003524163215761551, + "traffic_rms": 0.055206869881406315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0735210881639432e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.607321046389557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021160620678472717, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21970650088787078 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2628556489944458, + "mean_feedback_forward_cosine": 0.9856089218970268, + "mean_feedback_forward_relative_error": 0.16747105109595484, + "min_feedback_forward_cosine": 0.9662551879882812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034849494084121984, + "instruction_rms": 0.003484949407728848, + "raw_apical_rms": 0.05519303568909447, + "teaching_rms": 0.0034849494084121984, + "traffic_rms": 0.05508011596851206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0837832792608424e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.60587302112203e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02257218574662187, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21523030988693237 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2525036334991455, + "mean_feedback_forward_cosine": 0.9866824784586506, + "mean_feedback_forward_relative_error": 0.1610484448892455, + "min_feedback_forward_cosine": 0.9688282012939453 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034917662856806123, + "instruction_rms": 0.0034917662858501143, + "raw_apical_rms": 0.05527822677916193, + "teaching_rms": 0.0034917662856806123, + "traffic_rms": 0.05516500038946166 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.126600584517746e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6133050695794228e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02109727938863001, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.21525467616081237 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24268995225429535, + "mean_feedback_forward_cosine": 0.9876585891169887, + "mean_feedback_forward_relative_error": 0.15499278037778794, + "min_feedback_forward_cosine": 0.9711865186691284 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034075584107268234, + "instruction_rms": 0.0034075584107293444, + "raw_apical_rms": 0.05574660768048955, + "teaching_rms": 0.0034075584107268234, + "traffic_rms": 0.05563958960117571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1090711399551765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.611899712013731e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01983780699092181, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20698030591964722 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23288381099700928, + "mean_feedback_forward_cosine": 0.9885907403884395, + "mean_feedback_forward_relative_error": 0.1490106956372338, + "min_feedback_forward_cosine": 0.9734465479850769 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003478812216775989, + "instruction_rms": 0.0034788122157447727, + "raw_apical_rms": 0.0556278588216904, + "teaching_rms": 0.003478812216775989, + "traffic_rms": 0.05551554556490304 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.722580122986983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6100672064411973e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020488873967530775, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2107636794281006 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22301410138607025, + "mean_feedback_forward_cosine": 0.9894521620965773, + "mean_feedback_forward_relative_error": 0.14327095881585153, + "min_feedback_forward_cosine": 0.975633978843689 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034629061680117664, + "instruction_rms": 0.003462906169251398, + "raw_apical_rms": 0.05556501585941704, + "teaching_rms": 0.0034629061680117664, + "traffic_rms": 0.05545401907876476 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8279433078305374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6206850544553455e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021724748182863983, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.21416042904853821 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21437855064868927, + "mean_feedback_forward_cosine": 0.9902277723435433, + "mean_feedback_forward_relative_error": 0.1378802750860491, + "min_feedback_forward_cosine": 0.9774696826934814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003463676196754681, + "instruction_rms": 0.0034636761965173724, + "raw_apical_rms": 0.055525037407736684, + "teaching_rms": 0.003463676196754681, + "traffic_rms": 0.05541364836041376 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.7862751278130418e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6149845603655698e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020520803555521702, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20265906400680542 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20568503439426422, + "mean_feedback_forward_cosine": 0.9909787812540608, + "mean_feedback_forward_relative_error": 0.13247080135249323, + "min_feedback_forward_cosine": 0.9792330861091614 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003510927665270242, + "instruction_rms": 0.0035109276651843066, + "raw_apical_rms": 0.05582674945969147, + "teaching_rms": 0.003510927665270242, + "traffic_rms": 0.05571295206953726 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3045902963758635e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6118066149624304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02015539733433875, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.20840632899284361 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19723662734031677, + "mean_feedback_forward_cosine": 0.9916658516853086, + "mean_feedback_forward_relative_error": 0.1273198274354781, + "min_feedback_forward_cosine": 0.9808846712112427 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035062618602714713, + "instruction_rms": 0.003506261859345746, + "raw_apical_rms": 0.05615305845876284, + "teaching_rms": 0.0035062618602714713, + "traffic_rms": 0.05604037567921221 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.4129445108656e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6014736253142497e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02046214933652034, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.2021079085111618 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18926401436328888, + "mean_feedback_forward_cosine": 0.9923063881935612, + "mean_feedback_forward_relative_error": 0.12231636275687525, + "min_feedback_forward_cosine": 0.982383131980896 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003516765429870336, + "instruction_rms": 0.003516765430073125, + "raw_apical_rms": 0.05630377710605059, + "teaching_rms": 0.003516765429870336, + "traffic_rms": 0.05619054600727286 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.273791821494342e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6093534676097816e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02003378984931698, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.20293728838920594 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1818099468946457, + "mean_feedback_forward_cosine": 0.9929044285128193, + "mean_feedback_forward_relative_error": 0.11745534092187881, + "min_feedback_forward_cosine": 0.9837331771850586 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003539759094775681, + "instruction_rms": 0.0035397590935910806, + "raw_apical_rms": 0.05654376001571229, + "teaching_rms": 0.003539759094775681, + "traffic_rms": 0.05643005746843267 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1377569936475993e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6062807787320196e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02042036990613136, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.20467912701129914 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17471817135810852, + "mean_feedback_forward_cosine": 0.9934410664343065, + "mean_feedback_forward_relative_error": 0.1129209119225702, + "min_feedback_forward_cosine": 0.9849737882614136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035035772850481837, + "instruction_rms": 0.003503577285765434, + "raw_apical_rms": 0.05696834680452215, + "teaching_rms": 0.0035035772850481837, + "traffic_rms": 0.056857124350458955 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.077923078047661e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.600671472658845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02063863385998486, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.19943748482704163 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1675136387348175, + "mean_feedback_forward_cosine": 0.9939390863141706, + "mean_feedback_forward_relative_error": 0.10854704065188285, + "min_feedback_forward_cosine": 0.986180305480957 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003460854988457317, + "instruction_rms": 0.003460854987675644, + "raw_apical_rms": 0.056929572875802185, + "teaching_rms": 0.003460854988457317, + "traffic_rms": 0.05682131903857918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9052132316232928e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6032981285331634e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021201418376711866, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.20049855659961702 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16062714159488678, + "mean_feedback_forward_cosine": 0.994396315466973, + "mean_feedback_forward_relative_error": 0.10436200554813108, + "min_feedback_forward_cosine": 0.9872829914093018 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034394795283364727, + "instruction_rms": 0.0034394795282041484, + "raw_apical_rms": 0.05709322430710687, + "teaching_rms": 0.0034394795283364727, + "traffic_rms": 0.05698637815284925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.990803966123167e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5938515220044336e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022300192012534716, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1970031389427185 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15414825081825256, + "mean_feedback_forward_cosine": 0.99481027741586, + "mean_feedback_forward_relative_error": 0.1004205247567546, + "min_feedback_forward_cosine": 0.9882780313491821 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003407280902674854, + "instruction_rms": 0.0034072809029580134, + "raw_apical_rms": 0.05739876923595902, + "teaching_rms": 0.003407280902674854, + "traffic_rms": 0.05729407878094982 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.003417698404064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.599799140742495e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02125285705277914, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19356034287452697 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14795032143592834, + "mean_feedback_forward_cosine": 0.9951995841918453, + "mean_feedback_forward_relative_error": 0.09657560421093818, + "min_feedback_forward_cosine": 0.9891961216926575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034431538648207155, + "instruction_rms": 0.0034431538648933154, + "raw_apical_rms": 0.0575475442909583, + "teaching_rms": 0.0034431538648207155, + "traffic_rms": 0.05744136604177521 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.613916549876194e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6001467343181264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021523805420229147, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19540233695983886 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14202706515789032, + "mean_feedback_forward_cosine": 0.9955618535318682, + "mean_feedback_forward_relative_error": 0.09285408811223123, + "min_feedback_forward_cosine": 0.9900385141372681 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034527204156694496, + "instruction_rms": 0.0034527204153999356, + "raw_apical_rms": 0.0575860690935223, + "teaching_rms": 0.0034527204156694496, + "traffic_rms": 0.05747964931783119 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 4.097085692933433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5986994654430554e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01981468322199838, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1917427585887909 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13640093803405762, + "mean_feedback_forward_cosine": 0.9958990177800578, + "mean_feedback_forward_relative_error": 0.08924914163447195, + "min_feedback_forward_cosine": 0.9908096790313721 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034899244860362223, + "instruction_rms": 0.0034899244867008703, + "raw_apical_rms": 0.057695361379787986, + "teaching_rms": 0.0034899244860362223, + "traffic_rms": 0.05758637898625079 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8477079422041243e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.596245433932603e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019984007315012237, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18972671369552613 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13088887929916382, + "mean_feedback_forward_cosine": 0.9962043992934688, + "mean_feedback_forward_relative_error": 0.08585845396643685, + "min_feedback_forward_cosine": 0.9915297031402588 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034255154105749144, + "instruction_rms": 0.003425515409841862, + "raw_apical_rms": 0.05781750491073565, + "teaching_rms": 0.0034255154105749144, + "traffic_rms": 0.0577128162964886 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0972127146023922e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.593651062692368e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020320203604109334, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18884975323677064 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1257656067609787, + "mean_feedback_forward_cosine": 0.996486819559528, + "mean_feedback_forward_relative_error": 0.08259629265915963, + "min_feedback_forward_cosine": 0.9921789169311523 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034386235761980468, + "instruction_rms": 0.003438623575855904, + "raw_apical_rms": 0.05762120511331127, + "teaching_rms": 0.0034386235761980468, + "traffic_rms": 0.057515302598573514 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2180460135245994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.593734784510328e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020054449968256133, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18765089283466338 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12077493965625763, + "mean_feedback_forward_cosine": 0.9967486973731748, + "mean_feedback_forward_relative_error": 0.07945237513030728, + "min_feedback_forward_cosine": 0.9927830100059509 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003419543214946426, + "instruction_rms": 0.0034195432154761458, + "raw_apical_rms": 0.057684902107971385, + "teaching_rms": 0.003419543214946426, + "traffic_rms": 0.05758028212217341 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.94860643176753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.599778503908224e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021458118418434365, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18308923247337341 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1158987358212471, + "mean_feedback_forward_cosine": 0.996993476344693, + "mean_feedback_forward_relative_error": 0.07640130036780911, + "min_feedback_forward_cosine": 0.9933519959449768 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003423298513944789, + "instruction_rms": 0.0034232985145014267, + "raw_apical_rms": 0.05780217591358895, + "teaching_rms": 0.003423298513944789, + "traffic_rms": 0.05769765557000415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.6822231546930198e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6112852023474386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02141487857727527, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18563451991081237 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11102660000324249, + "mean_feedback_forward_cosine": 0.9972287666413092, + "mean_feedback_forward_relative_error": 0.0733463716122412, + "min_feedback_forward_cosine": 0.9938979744911194 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035001837517003036, + "instruction_rms": 0.00350018375225057, + "raw_apical_rms": 0.05781039268734066, + "teaching_rms": 0.0035001837517003036, + "traffic_rms": 0.057700930404744825 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.851858222323699e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6008388415764536e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02140994021673855, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.19004285793304443 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10672832280397415, + "mean_feedback_forward_cosine": 0.9974359446956266, + "mean_feedback_forward_relative_error": 0.07053910412134663, + "min_feedback_forward_cosine": 0.9943585395812988 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034686207722556942, + "instruction_rms": 0.0034686207723892844, + "raw_apical_rms": 0.05797376054847445, + "teaching_rms": 0.0034686207722556942, + "traffic_rms": 0.0578663888368281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1717069031979008e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.595235618123169e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020836060092768772, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.1814848718738556 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10235501080751419, + "mean_feedback_forward_cosine": 0.9976284907710168, + "mean_feedback_forward_relative_error": 0.06784144632758633, + "min_feedback_forward_cosine": 0.9948086142539978 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033951238837711677, + "instruction_rms": 0.00339512388331119, + "raw_apical_rms": 0.05817950912657142, + "teaching_rms": 0.0033951238837711677, + "traffic_rms": 0.05807703550395544 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.753015865801899e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.600331763892189e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020213499625784597, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.18406702889442444 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09815677255392075, + "mean_feedback_forward_cosine": 0.9978054031249015, + "mean_feedback_forward_relative_error": 0.06526561213597175, + "min_feedback_forward_cosine": 0.9952248334884644 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033916490041799594, + "instruction_rms": 0.003391649003639008, + "raw_apical_rms": 0.0584052550468733, + "teaching_rms": 0.0033916490041799594, + "traffic_rms": 0.058303854638573474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.122481035371493e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.598092720213753e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01965742485662164, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1818651003932953 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09418652206659317, + "mean_feedback_forward_cosine": 0.9979652550912672, + "mean_feedback_forward_relative_error": 0.06283595869618078, + "min_feedback_forward_cosine": 0.9956014156341553 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003393359881132555, + "instruction_rms": 0.0033933598807929528, + "raw_apical_rms": 0.05839581313424848, + "teaching_rms": 0.003393359881132555, + "traffic_rms": 0.05829397508100326 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8277688695889083e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.601090035300354e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01935683600017686, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.18304167880535127 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09051638096570969, + "mean_feedback_forward_cosine": 0.9981175257313636, + "mean_feedback_forward_relative_error": 0.06043314320906516, + "min_feedback_forward_cosine": 0.9959362745285034 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003421384726465541, + "instruction_rms": 0.0034213847260471217, + "raw_apical_rms": 0.05860390225111553, + "teaching_rms": 0.003421384726465541, + "traffic_rms": 0.0585005453254147 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1704293473590042e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6052442861151937e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020665724623899505, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17877166667938232 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0869675725698471, + "mean_feedback_forward_cosine": 0.9982598962322358, + "mean_feedback_forward_relative_error": 0.05810259396751081, + "min_feedback_forward_cosine": 0.9962478876113892 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003430392736216952, + "instruction_rms": 0.0034303927350730185, + "raw_apical_rms": 0.058734110358252746, + "teaching_rms": 0.003430392736216952, + "traffic_rms": 0.05863018095070694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.926413517594483e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5858016364409023e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019182674413133743, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17750683407783507 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08344399929046631, + "mean_feedback_forward_cosine": 0.9983924300439896, + "mean_feedback_forward_relative_error": 0.05584745854139328, + "min_feedback_forward_cosine": 0.9965448379516602 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003456510816962367, + "instruction_rms": 0.0034565108175854072, + "raw_apical_rms": 0.05886184411149324, + "teaching_rms": 0.003456510816962367, + "traffic_rms": 0.058757089948110235 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.8557835207720927e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6042757269500776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02055267449983377, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.18042453233718872 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08004778623580933, + "mean_feedback_forward_cosine": 0.9985132544271408, + "mean_feedback_forward_relative_error": 0.05370641517783365, + "min_feedback_forward_cosine": 0.9968206286430359 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034141354764852523, + "instruction_rms": 0.0034141354754085615, + "raw_apical_rms": 0.058834754110088996, + "teaching_rms": 0.0034141354764852523, + "traffic_rms": 0.058732339688646315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0750956508663876e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589604463812503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02017676912619823, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17927248138427734 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07687560468912125, + "mean_feedback_forward_cosine": 0.9986259322012624, + "mean_feedback_forward_relative_error": 0.05162628139218976, + "min_feedback_forward_cosine": 0.9970664381980896 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034341814803397477, + "instruction_rms": 0.0034341814810828794, + "raw_apical_rms": 0.05889412220055457, + "teaching_rms": 0.0034341814803397477, + "traffic_rms": 0.058790399553382916 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.979074859548291e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.605374611941103e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019630663659274142, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.17652707487344743 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07375805824995041, + "mean_feedback_forward_cosine": 0.9987325053061208, + "mean_feedback_forward_relative_error": 0.049586879209645336, + "min_feedback_forward_cosine": 0.9972988963127136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003475087451530602, + "instruction_rms": 0.0034750874520372422, + "raw_apical_rms": 0.0589721955658991, + "teaching_rms": 0.003475087451530602, + "traffic_rms": 0.0588657175678156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2401831500928893e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5850199399309963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019999401937577193, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17692116313934325 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707392767071724, + "mean_feedback_forward_cosine": 0.998829476294979, + "mean_feedback_forward_relative_error": 0.04765409141057922, + "min_feedback_forward_cosine": 0.9975142478942871 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003445978700431747, + "instruction_rms": 0.0034459787004751346, + "raw_apical_rms": 0.05886002576594397, + "teaching_rms": 0.003445978700431747, + "traffic_rms": 0.058755857332201906 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0091544428169072e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5963862271862235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019010066002730936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17988833545684815 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06793131679296494, + "mean_feedback_forward_cosine": 0.9989166490493282, + "mean_feedback_forward_relative_error": 0.04584274152594228, + "min_feedback_forward_cosine": 0.9977070093154907 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00341759967618036, + "instruction_rms": 0.003417599675177702, + "raw_apical_rms": 0.05923595540794477, + "teaching_rms": 0.00341759967618036, + "traffic_rms": 0.05913389437091155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2458282344928193e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5976921910014363e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020792721901837022, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1772440785264969 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06527094542980194, + "mean_feedback_forward_cosine": 0.9989996437103518, + "mean_feedback_forward_relative_error": 0.04404804362885414, + "min_feedback_forward_cosine": 0.9978823661804199 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003458292265493477, + "instruction_rms": 0.003458292265283366, + "raw_apical_rms": 0.05946490160713266, + "teaching_rms": 0.003458292265493477, + "traffic_rms": 0.05936067228041016 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.7556333276047553e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5942433066091953e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02065062018500329, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.17513171265602112 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06267433613538742, + "mean_feedback_forward_cosine": 0.9990758299827576, + "mean_feedback_forward_relative_error": 0.042338665664917036, + "min_feedback_forward_cosine": 0.9980476498603821 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003463749464852256, + "instruction_rms": 0.003463749466015367, + "raw_apical_rms": 0.059668149363479576, + "teaching_rms": 0.003463749464852256, + "traffic_rms": 0.05956380177314484 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0572395342810523e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5871959618824853e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02020075408532083, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.17257637537956239 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06021619588136673, + "mean_feedback_forward_cosine": 0.9991444983789998, + "mean_feedback_forward_relative_error": 0.04073551874006948, + "min_feedback_forward_cosine": 0.9981976747512817 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034053513403415818, + "instruction_rms": 0.003405351340818926, + "raw_apical_rms": 0.05981797354440699, + "teaching_rms": 0.0034053513403415818, + "traffic_rms": 0.05971833891337273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0854244553447643e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585660915558724e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02235868920569362, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.17234158856868745 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.057775773108005524, + "mean_feedback_forward_cosine": 0.999208927154541, + "mean_feedback_forward_relative_error": 0.03916807100176811, + "min_feedback_forward_cosine": 0.9983402490615845 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003438324558695901, + "instruction_rms": 0.003438324559016137, + "raw_apical_rms": 0.05982131544124343, + "teaching_rms": 0.003438324558695901, + "traffic_rms": 0.059718879839451615 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.205450577719148e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5898127222863757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021763737118733202, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.17341863000631333 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05547836422920227, + "mean_feedback_forward_cosine": 0.9992696604421062, + "mean_feedback_forward_relative_error": 0.03763908841797421, + "min_feedback_forward_cosine": 0.9984691739082336 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033998637072948225, + "instruction_rms": 0.0033998637077746106, + "raw_apical_rms": 0.059891200533465784, + "teaching_rms": 0.0033998637072948225, + "traffic_rms": 0.05979121869753013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.348518535200128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.593869182762272e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022518478104180433, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1750432247209549 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.053380243480205536, + "mean_feedback_forward_cosine": 0.9993224432391505, + "mean_feedback_forward_relative_error": 0.03624790047685946, + "min_feedback_forward_cosine": 0.9985824823379517 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033250119537614757, + "instruction_rms": 0.003325011954065987, + "raw_apical_rms": 0.05999480491245881, + "teaching_rms": 0.0033250119537614757, + "traffic_rms": 0.05989926423755988 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.6011905518430467e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.586319604544755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020896972466268936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1650091945362091 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051124557852745056, + "mean_feedback_forward_cosine": 0.9993754663775044, + "mean_feedback_forward_relative_error": 0.034804713401582935, + "min_feedback_forward_cosine": 0.9986997246742249 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003439834988753, + "instruction_rms": 0.0034398349874303394, + "raw_apical_rms": 0.059864987994036986, + "teaching_rms": 0.003439834988753, + "traffic_rms": 0.05976244116146561 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0462804783960564e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.595384501116954e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021231430526580674, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.17371611179351806 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.049127962440252304, + "mean_feedback_forward_cosine": 0.9994224559876227, + "mean_feedback_forward_relative_error": 0.03346922768339995, + "min_feedback_forward_cosine": 0.9987989664077759 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033619010933833866, + "instruction_rms": 0.0033619010927225606, + "raw_apical_rms": 0.059982092012269235, + "teaching_rms": 0.0033619010933833866, + "traffic_rms": 0.059884689203573024 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5894571820117562e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5867694896197485e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02262959573934223, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1715626421737671 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04728499427437782, + "mean_feedback_forward_cosine": 0.9994647118353075, + "mean_feedback_forward_relative_error": 0.03222058653350799, + "min_feedback_forward_cosine": 0.9988870620727539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033548353918590227, + "instruction_rms": 0.003354835392269414, + "raw_apical_rms": 0.06021775997509343, + "teaching_rms": 0.0033548353918590227, + "traffic_rms": 0.06012094352483448 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1354871648782137e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5942450127443865e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021305525827439938, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16552210284233093 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0453660674393177, + "mean_feedback_forward_cosine": 0.9995052737574424, + "mean_feedback_forward_relative_error": 0.030979650485659798, + "min_feedback_forward_cosine": 0.9989755153656006 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033726637883326003, + "instruction_rms": 0.003372663788674758, + "raw_apical_rms": 0.06029498339469778, + "teaching_rms": 0.0033726637883326003, + "traffic_rms": 0.06019784185951271 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0367517450381456e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591734790050541e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02159472278103336, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.166471966714859 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04361632093787193, + "mean_feedback_forward_cosine": 0.9995416075952591, + "mean_feedback_forward_relative_error": 0.0298194095131851, + "min_feedback_forward_cosine": 0.999052882194519 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003350492698055595, + "instruction_rms": 0.0033504926972128283, + "raw_apical_rms": 0.060371947030863606, + "teaching_rms": 0.003350492698055595, + "traffic_rms": 0.06027512436461606 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9863561462329926e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5909891750943932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02222160719281142, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.16222234501838684 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0418466217815876, + "mean_feedback_forward_cosine": 0.9995770569770567, + "mean_feedback_forward_relative_error": 0.028643882923549222, + "min_feedback_forward_cosine": 0.9991282224655151 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003402920333988259, + "instruction_rms": 0.0034029203338786632, + "raw_apical_rms": 0.060599267349975665, + "teaching_rms": 0.003402920333988259, + "traffic_rms": 0.06050043041849693 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1999248644988256e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5920537855278048e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021634372300728663, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.16844359785795213 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.040160052478313446, + "mean_feedback_forward_cosine": 0.9996090704394925, + "mean_feedback_forward_relative_error": 0.027540143879671252, + "min_feedback_forward_cosine": 0.9991968274116516 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003356290152885211, + "instruction_rms": 0.0033562901526771128, + "raw_apical_rms": 0.06075698200092619, + "teaching_rms": 0.003356290152885211, + "traffic_rms": 0.06066119929771656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.791095805003806e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5871488683198588e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021403321731415747, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.16844806903362275 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03851350024342537, + "mean_feedback_forward_cosine": 0.9996402340550576, + "mean_feedback_forward_relative_error": 0.026423181977964217, + "min_feedback_forward_cosine": 0.9992611408233643 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034480014756884305, + "instruction_rms": 0.003448001476853413, + "raw_apical_rms": 0.060677221645840704, + "teaching_rms": 0.0034480014756884305, + "traffic_rms": 0.060575920474552736 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0714656189084053, + "maximum_absolute_post_projection_soma_slope": 2.2089638562761138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5956784699897934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022100300203414663, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1734725644016266 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03697704151272774, + "mean_feedback_forward_cosine": 0.999667344554778, + "mean_feedback_forward_relative_error": 0.025405359274197008, + "min_feedback_forward_cosine": 0.9993188381195068 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033675156976960725, + "instruction_rms": 0.003367515698361574, + "raw_apical_rms": 0.06069001300979408, + "teaching_rms": 0.0033675156976960725, + "traffic_rms": 0.06059314659661246 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8590840866750113e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5938492771856475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021644090576521714, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16188694938659667 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03550988435745239, + "mean_feedback_forward_cosine": 0.999692557319518, + "mean_feedback_forward_relative_error": 0.024426692345690344, + "min_feedback_forward_cosine": 0.9993716478347778 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003390164787643623, + "instruction_rms": 0.0033901647879523524, + "raw_apical_rms": 0.060777765106341426, + "teaching_rms": 0.003390164787643623, + "traffic_rms": 0.06068032095663359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.8044050637276996e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.586254664389376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0215965019604694, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1648994934654236 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.034106262028217316, + "mean_feedback_forward_cosine": 0.9997149859705279, + "mean_feedback_forward_relative_error": 0.023516869833392483, + "min_feedback_forward_cosine": 0.9994202256202698 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003316037417653443, + "instruction_rms": 0.0033160374185728725, + "raw_apical_rms": 0.060605500052847226, + "teaching_rms": 0.003316037417653443, + "traffic_rms": 0.06051163708531536 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5912299861374777e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589362477194766e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02295299882431037, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16531682719707488 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03278369829058647, + "mean_feedback_forward_cosine": 0.999736359042506, + "mean_feedback_forward_relative_error": 0.022617473747701414, + "min_feedback_forward_cosine": 0.9994643926620483 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033429628725576794, + "instruction_rms": 0.0033429628718452888, + "raw_apical_rms": 0.060815451593235766, + "teaching_rms": 0.0033429628725576794, + "traffic_rms": 0.0607199330047135 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.8780577093480133e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5905770562756898e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022397742168927814, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16122076577186584 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03153737634420395, + "mean_feedback_forward_cosine": 0.9997553748469199, + "mean_feedback_forward_relative_error": 0.021786171221925367, + "min_feedback_forward_cosine": 0.9995043277740479 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0032894955815286554, + "instruction_rms": 0.003289495581533522, + "raw_apical_rms": 0.060982174378095014, + "teaching_rms": 0.0032894955815286554, + "traffic_rms": 0.06089030213465361 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.830583329365254e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5912228043044425e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021925054040415594, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.16027390190124513 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03152688220143318, + "mean_feedback_forward_cosine": 0.9997555440472018, + "mean_feedback_forward_relative_error": 0.021776764563495112, + "min_feedback_forward_cosine": 0.999504566192627 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002398621889958721, + "instruction_rms": 0.0023986218896793515, + "raw_apical_rms": 0.06104121477701332, + "teaching_rms": 0.002398621889958721, + "traffic_rms": 0.06099145334045525 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1113459425237124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5627137032607413e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01982363357403087, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.09333422934532165 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03152167797088623, + "mean_feedback_forward_cosine": 0.999755653642839, + "mean_feedback_forward_relative_error": 0.02177059085619065, + "min_feedback_forward_cosine": 0.9995048642158508 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021333371362211224, + "instruction_rms": 0.00213333713588387, + "raw_apical_rms": 0.06096595887695236, + "teaching_rms": 0.0021333371362211224, + "traffic_rms": 0.06092623399803428 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2154578616095932e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5513763268625097e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020220379507792798, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06759175801753997 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03151693195104599, + "mean_feedback_forward_cosine": 0.9997557401657104, + "mean_feedback_forward_relative_error": 0.021764965367413337, + "min_feedback_forward_cosine": 0.9995049238204956 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002046703623203282, + "instruction_rms": 0.00204670362326996, + "raw_apical_rms": 0.060937144381639934, + "teaching_rms": 0.002046703623203282, + "traffic_rms": 0.06090011029269463 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.957503670269034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556313958042445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02042894344610579, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.05698678456664085 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031512632966041565, + "mean_feedback_forward_cosine": 0.99975581899766, + "mean_feedback_forward_relative_error": 0.02175972382387807, + "min_feedback_forward_cosine": 0.9995051026344299 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019799681669607806, + "instruction_rms": 0.001979968166781406, + "raw_apical_rms": 0.06087065603522165, + "teaching_rms": 0.0019799681669607806, + "traffic_rms": 0.06083562300761488 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3360593459642587e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5549141117006324e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020541968774159366, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.05106272052049637 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03150852397084236, + "mean_feedback_forward_cosine": 0.9997559055205314, + "mean_feedback_forward_relative_error": 0.021754666847447233, + "min_feedback_forward_cosine": 0.9995052218437195 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019290545186167257, + "instruction_rms": 0.0019290545190684039, + "raw_apical_rms": 0.060821752086465654, + "teaching_rms": 0.0019290545186167257, + "traffic_rms": 0.06078820523870638 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3436090401673937e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563156615255852e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020825996781660347, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.045936097958087924 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03150397539138794, + "mean_feedback_forward_cosine": 0.9997559805070201, + "mean_feedback_forward_relative_error": 0.02174961195898152, + "min_feedback_forward_cosine": 0.9995054006576538 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019731081252130625, + "instruction_rms": 0.0019731081252151576, + "raw_apical_rms": 0.060766449778849665, + "teaching_rms": 0.0019731081252130625, + "traffic_rms": 0.06073123922860457 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2527224530222156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5606805083342983e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02092268456358297, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04352586357593537 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03149962052702904, + "mean_feedback_forward_cosine": 0.9997560554935087, + "mean_feedback_forward_relative_error": 0.021744657217735244, + "min_feedback_forward_cosine": 0.9995054006576538 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019308048434830663, + "instruction_rms": 0.0019308048438191094, + "raw_apical_rms": 0.060719777214897394, + "teaching_rms": 0.0019308048434830663, + "traffic_rms": 0.06068535056216789 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2302476310187558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5591246223633387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021232555423782, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.0400309817135334 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0314955972135067, + "mean_feedback_forward_cosine": 0.9997561477845714, + "mean_feedback_forward_relative_error": 0.02173995174046966, + "min_feedback_forward_cosine": 0.9995056390762329 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018817973347551712, + "instruction_rms": 0.0018817973349551723, + "raw_apical_rms": 0.06068903866925442, + "teaching_rms": 0.0018817973347551712, + "traffic_rms": 0.060656312031160925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3623050182663974e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.565024513138787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02115694024026615, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.037535460599660876 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03149155154824257, + "mean_feedback_forward_cosine": 0.9997562246937906, + "mean_feedback_forward_relative_error": 0.021735391474419063, + "min_feedback_forward_cosine": 0.9995057582855225 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018373676320763686, + "instruction_rms": 0.0018373676323064534, + "raw_apical_rms": 0.06064975821286218, + "teaching_rms": 0.0018373676320763686, + "traffic_rms": 0.060618067379224747 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.086065009621052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5639751893530308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020964269746619, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.0344030801987648 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03148741275072098, + "mean_feedback_forward_cosine": 0.9997563016030097, + "mean_feedback_forward_relative_error": 0.021730675872775815, + "min_feedback_forward_cosine": 0.9995057582855225 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001859365300345051, + "instruction_rms": 0.0018593653007709069, + "raw_apical_rms": 0.06063491070046303, + "teaching_rms": 0.001859365300345051, + "traffic_rms": 0.060601999199207876 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0311309967269153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5645318901013604e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021201538473130686, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03404332453489303 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03148353099822998, + "mean_feedback_forward_cosine": 0.9997563708213068, + "mean_feedback_forward_relative_error": 0.021726243648557894, + "min_feedback_forward_cosine": 0.9995059967041016 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017792693721633372, + "instruction_rms": 0.001779269371810353, + "raw_apical_rms": 0.06060974513709711, + "teaching_rms": 0.0017792693721633372, + "traffic_rms": 0.0605800890919125 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.143865351911245e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562937477637461e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021216626116228588, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.030716377469301222 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03147963806986809, + "mean_feedback_forward_cosine": 0.9997564515759868, + "mean_feedback_forward_relative_error": 0.021721871239283392, + "min_feedback_forward_cosine": 0.9995061159133911 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001778277828207509, + "instruction_rms": 0.001778277828723965, + "raw_apical_rms": 0.06057176292817008, + "teaching_rms": 0.001778277828207509, + "traffic_rms": 0.06054181141671 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3181401687111247e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5729326947873305e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021322696570138612, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.029599104360342025 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03147570416331291, + "mean_feedback_forward_cosine": 0.9997565438670497, + "mean_feedback_forward_relative_error": 0.021717479452490807, + "min_feedback_forward_cosine": 0.9995062351226807 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018072599197181153, + "instruction_rms": 0.001807259919694294, + "raw_apical_rms": 0.060517731868169854, + "teaching_rms": 0.0018072599197181153, + "traffic_rms": 0.06048620141870423 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1655408133369747e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670435557415783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021250775148907254, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.02921918411910534 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031471576541662216, + "mean_feedback_forward_cosine": 0.9997566150080773, + "mean_feedback_forward_relative_error": 0.021713021166262128, + "min_feedback_forward_cosine": 0.999506413936615 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017855020169386564, + "instruction_rms": 0.001785502017071911, + "raw_apical_rms": 0.06049119120247394, + "teaching_rms": 0.0017855020169386564, + "traffic_rms": 0.06045980743748742 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9960024744136717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5606584947692495e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021636498446608168, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02790903121113777 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03146752715110779, + "mean_feedback_forward_cosine": 0.9997566996082183, + "mean_feedback_forward_relative_error": 0.02170853429443894, + "min_feedback_forward_cosine": 0.9995065331459045 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017775455003777672, + "instruction_rms": 0.0017775455005886673, + "raw_apical_rms": 0.06051045129667271, + "teaching_rms": 0.0017775455003777672, + "traffic_rms": 0.06047985973341938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8998475681542004e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5666541372563246e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021366600157939744, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.025734815603494644 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031463682651519775, + "mean_feedback_forward_cosine": 0.9997567688265154, + "mean_feedback_forward_relative_error": 0.021704253605416706, + "min_feedback_forward_cosine": 0.9995065927505493 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017261972222227678, + "instruction_rms": 0.0017261972216782895, + "raw_apical_rms": 0.06048572644802996, + "teaching_rms": 0.0017261972222227678, + "traffic_rms": 0.06045675628127358 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.436298984775931e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5655378692333757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021085676031754666, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024729185898303984 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031459592282772064, + "mean_feedback_forward_cosine": 0.9997568572721174, + "mean_feedback_forward_relative_error": 0.02169988788063488, + "min_feedback_forward_cosine": 0.9995068311691284 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017572288449339136, + "instruction_rms": 0.0017572288450132649, + "raw_apical_rms": 0.06046964119418366, + "teaching_rms": 0.0017572288449339136, + "traffic_rms": 0.060438826447286945 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.188399506053429e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.572685545345221e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021047060669153846, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.025334170320630074 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03145581856369972, + "mean_feedback_forward_cosine": 0.999756932258606, + "mean_feedback_forward_relative_error": 0.02169556271344904, + "min_feedback_forward_cosine": 0.999506950378418 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001701262973061038, + "instruction_rms": 0.0017012629733924403, + "raw_apical_rms": 0.06043503368079829, + "teaching_rms": 0.001701262973061038, + "traffic_rms": 0.0604062522732447 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.025842071873285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5722834470548688e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021035353204275214, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.022988191098570823 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03145194798707962, + "mean_feedback_forward_cosine": 0.9997570187814774, + "mean_feedback_forward_relative_error": 0.021691336356584105, + "min_feedback_forward_cosine": 0.9995070099830627 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001680353697396918, + "instruction_rms": 0.0016803536968119162, + "raw_apical_rms": 0.0604305457404473, + "teaching_rms": 0.001680353697396918, + "traffic_rms": 0.06040247182631898 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.249623420880198e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5615927038061534e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02142448054117058, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.02194026156425476 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03144793212413788, + "mean_feedback_forward_cosine": 0.9997571129952708, + "mean_feedback_forward_relative_error": 0.021687014809539242, + "min_feedback_forward_cosine": 0.9995071291923523 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001710373941006606, + "instruction_rms": 0.0017103739407916924, + "raw_apical_rms": 0.06044723920612104, + "teaching_rms": 0.001710373941006606, + "traffic_rms": 0.060417783930974536 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.4688869615374642e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563692249549339e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021262344263362223, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.02244715241879225 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031444188207387924, + "mean_feedback_forward_cosine": 0.9997571822135679, + "mean_feedback_forward_relative_error": 0.02168291550309908, + "min_feedback_forward_cosine": 0.9995072484016418 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016376603236343352, + "instruction_rms": 0.0016376603239503544, + "raw_apical_rms": 0.0604257899673085, + "teaching_rms": 0.0016376603236343352, + "traffic_rms": 0.06039848206828698 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.129364418124169e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5641752863823337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021192157258718636, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.02022576591670513 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0314406156539917, + "mean_feedback_forward_cosine": 0.9997572283590993, + "mean_feedback_forward_relative_error": 0.021678831368204084, + "min_feedback_forward_cosine": 0.9995072484016418 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015659922530151065, + "instruction_rms": 0.0015659922533562884, + "raw_apical_rms": 0.06038533492717041, + "teaching_rms": 0.0015659922530151065, + "traffic_rms": 0.060359866341804125 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.7033721039515513e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670427066009977e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021187531568565437, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01860323529779911 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03143704682588577, + "mean_feedback_forward_cosine": 0.9997573168047013, + "mean_feedback_forward_relative_error": 0.021674805425948674, + "min_feedback_forward_cosine": 0.9995075464248657 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001558712258865541, + "instruction_rms": 0.0015587122583099176, + "raw_apical_rms": 0.06033626783480335, + "teaching_rms": 0.001558712258865541, + "traffic_rms": 0.06031083462316474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2601041038683434e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558320617213378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021261311529394988, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.01847828521847725 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03143325075507164, + "mean_feedback_forward_cosine": 0.9997574014048423, + "mean_feedback_forward_relative_error": 0.02167059356466897, + "min_feedback_forward_cosine": 0.9995076060295105 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016103571251713815, + "instruction_rms": 0.0016103571255068685, + "raw_apical_rms": 0.06032843384180606, + "teaching_rms": 0.0016103571251713815, + "traffic_rms": 0.060301178429157186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.978660162966662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5696847499502945e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021231445869597243, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.019125866518616676 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03142950311303139, + "mean_feedback_forward_cosine": 0.9997574860049833, + "mean_feedback_forward_relative_error": 0.02166658192272148, + "min_feedback_forward_cosine": 0.9995076656341553 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015375907247758183, + "instruction_rms": 0.0015375907247493494, + "raw_apical_rms": 0.06031516221230108, + "teaching_rms": 0.0015375907247758183, + "traffic_rms": 0.060289711654303434 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3115177327781566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5771107900153756e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02155187387803211, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.017221306319236754 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031425733119249344, + "mean_feedback_forward_cosine": 0.999757549455089, + "mean_feedback_forward_relative_error": 0.02166240917698991, + "min_feedback_forward_cosine": 0.9995078444480896 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015801354373825949, + "instruction_rms": 0.0015801354367632253, + "raw_apical_rms": 0.060280644505303904, + "teaching_rms": 0.0015801354373825949, + "traffic_rms": 0.06025394289452608 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.175109514723772e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5608545608303522e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021750289061752293, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.017284111934304236 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03142174705862999, + "mean_feedback_forward_cosine": 0.9997576513598042, + "mean_feedback_forward_relative_error": 0.02165804100373099, + "min_feedback_forward_cosine": 0.9995079040527344 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001617392017166664, + "instruction_rms": 0.0016173920171565102, + "raw_apical_rms": 0.06029715769635445, + "teaching_rms": 0.001617392017166664, + "traffic_rms": 0.0602691789405435 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.3002456828180584e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.564494487473074e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021892627999324368, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.0171700565508008 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031418368220329285, + "mean_feedback_forward_cosine": 0.9997577090417186, + "mean_feedback_forward_relative_error": 0.021654212426754735, + "min_feedback_forward_cosine": 0.9995079636573792 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014468703310447336, + "instruction_rms": 0.0014468703307280186, + "raw_apical_rms": 0.06029150587431078, + "teaching_rms": 0.0014468703310447336, + "traffic_rms": 0.06026915762483013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.9801536410568588e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568471699537214e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02102892052471611, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01475954680517316 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03141402825713158, + "mean_feedback_forward_cosine": 0.9997578032555119, + "mean_feedback_forward_relative_error": 0.02164962124680319, + "min_feedback_forward_cosine": 0.999508261680603 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001647078165995079, + "instruction_rms": 0.0016470781654027529, + "raw_apical_rms": 0.060317028455909104, + "teaching_rms": 0.001647078165995079, + "traffic_rms": 0.06028755363885058 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8321864914128128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5684968965499933e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021455415856616818, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.016771517105400563 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03141002729535103, + "mean_feedback_forward_cosine": 0.9997578782420005, + "mean_feedback_forward_relative_error": 0.021645333362563003, + "min_feedback_forward_cosine": 0.9995083212852478 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015545478188276319, + "instruction_rms": 0.0015545478188123511, + "raw_apical_rms": 0.060325860883568796, + "teaching_rms": 0.0015545478188276319, + "traffic_rms": 0.06029907231392622 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9989645494433717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5680279048596704e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021094721926493518, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.01596706278845668 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03140585124492645, + "mean_feedback_forward_cosine": 0.9997579686103328, + "mean_feedback_forward_relative_error": 0.021641116859691757, + "min_feedback_forward_cosine": 0.9995084404945374 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015275423809823635, + "instruction_rms": 0.0015275423811523228, + "raw_apical_rms": 0.06032168137533785, + "teaching_rms": 0.0015275423809823635, + "traffic_rms": 0.06029643454098172 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.002989063498717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570085999278273e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021163578902115926, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.016134593434929848 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031401731073856354, + "mean_feedback_forward_cosine": 0.9997580493650129, + "mean_feedback_forward_relative_error": 0.02163685822198468, + "min_feedback_forward_cosine": 0.9995085000991821 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015095788660353675, + "instruction_rms": 0.0015095788659874414, + "raw_apical_rms": 0.06029600621394194, + "teaching_rms": 0.0015095788660353675, + "traffic_rms": 0.06027074915240652 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1048956355684822e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5688137019403565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021225011742469265, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.014972648302912712 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031398188322782516, + "mean_feedback_forward_cosine": 0.999758130119693, + "mean_feedback_forward_relative_error": 0.021632933974145882, + "min_feedback_forward_cosine": 0.9995087385177612 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00141278160256858, + "instruction_rms": 0.0014127816022092617, + "raw_apical_rms": 0.060307643128228026, + "teaching_rms": 0.00141278160256858, + "traffic_rms": 0.06028470445843777 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2203876071102968e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5596273518653564e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021440215175471164, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.013288077376578004 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03139420598745346, + "mean_feedback_forward_cosine": 0.9997582070289119, + "mean_feedback_forward_relative_error": 0.021628684168981928, + "min_feedback_forward_cosine": 0.9995088577270508 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014683729445171235, + "instruction_rms": 0.001468372944926815, + "raw_apical_rms": 0.060272037375291125, + "teaching_rms": 0.0014683729445171235, + "traffic_rms": 0.06024697004198889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.237959684237012e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5709243206294097e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021676522310623222, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.014023889936506748 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313900001347065, + "mean_feedback_forward_cosine": 0.999758287783592, + "mean_feedback_forward_relative_error": 0.02162433272948669, + "min_feedback_forward_cosine": 0.9995089769363403 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00151604943175004, + "instruction_rms": 0.0015160494313732247, + "raw_apical_rms": 0.06026722077950062, + "teaching_rms": 0.00151604943175004, + "traffic_rms": 0.06024168792251169 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5872678222071954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568476729291428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021301102569107767, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.013910914593338966 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031386032700538635, + "mean_feedback_forward_cosine": 0.999758376229194, + "mean_feedback_forward_relative_error": 0.021620102587246125, + "min_feedback_forward_cosine": 0.9995090365409851 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001477430439914596, + "instruction_rms": 0.0014774304402341402, + "raw_apical_rms": 0.06029351024500925, + "teaching_rms": 0.001477430439914596, + "traffic_rms": 0.06026928877265612 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0881808282524617e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5635215935810226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02103406847482763, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.013774006176888943 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03138156980276108, + "mean_feedback_forward_cosine": 0.9997584608293348, + "mean_feedback_forward_relative_error": 0.021615536868452065, + "min_feedback_forward_cosine": 0.9995092153549194 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015153848135794515, + "instruction_rms": 0.0015153848135075166, + "raw_apical_rms": 0.06028979729879795, + "teaching_rms": 0.0015153848135794515, + "traffic_rms": 0.060263449959591736 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0386327292953865e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570405912992581e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02128528778893488, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.014216195790618658 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03137783706188202, + "mean_feedback_forward_cosine": 0.9997585242794406, + "mean_feedback_forward_relative_error": 0.021611571477185332, + "min_feedback_forward_cosine": 0.9995093941688538 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013623057297024645, + "instruction_rms": 0.0013623057298733391, + "raw_apical_rms": 0.06026149093538735, + "teaching_rms": 0.0013623057297024645, + "traffic_rms": 0.060239669731314074 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.876320077087712e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5702227385891922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021722423490932816, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012675859419107437 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031373877078294754, + "mean_feedback_forward_cosine": 0.9997586242614254, + "mean_feedback_forward_relative_error": 0.02160714517137216, + "min_feedback_forward_cosine": 0.9995095133781433 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014160875186484813, + "instruction_rms": 0.0014160875188215937, + "raw_apical_rms": 0.060271893141315994, + "teaching_rms": 0.0014160875186484813, + "traffic_rms": 0.06024855283609319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1365448077403926e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5707145004668614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02166383299546484, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012604654445797205 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03137008845806122, + "mean_feedback_forward_cosine": 0.999758691556992, + "mean_feedback_forward_relative_error": 0.021602971178870046, + "min_feedback_forward_cosine": 0.9995096325874329 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014017556826610889, + "instruction_rms": 0.0014017556817953554, + "raw_apical_rms": 0.06021674431885879, + "teaching_rms": 0.0014017556826610889, + "traffic_rms": 0.06019406225614311 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.13758344358439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5693120693005157e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02144647719055143, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012337776693105698 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03136617690324783, + "mean_feedback_forward_cosine": 0.9997587934617074, + "mean_feedback_forward_relative_error": 0.021598879428159808, + "min_feedback_forward_cosine": 0.9995098114013672 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001359620341620725, + "instruction_rms": 0.0013596203418669865, + "raw_apical_rms": 0.06023319874960461, + "teaching_rms": 0.001359620341620725, + "traffic_rms": 0.060211836187975896 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9645478133156757e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5742544068754367e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021716134699330095, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.0115347054579854 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031361497938632965, + "mean_feedback_forward_cosine": 0.9997588857527702, + "mean_feedback_forward_relative_error": 0.02159409573481929, + "min_feedback_forward_cosine": 0.9995098114013672 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015325011894380678, + "instruction_rms": 0.0015325011897909257, + "raw_apical_rms": 0.0602282827106931, + "teaching_rms": 0.0015325011894380678, + "traffic_rms": 0.060200925790311034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.6773845362981774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5832496708826057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02185208231985076, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.013095144706964493 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313575379550457, + "mean_feedback_forward_cosine": 0.9997589703529112, + "mean_feedback_forward_relative_error": 0.021589957237724337, + "min_feedback_forward_cosine": 0.9995100498199463 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013291754192877106, + "instruction_rms": 0.0013291754186445942, + "raw_apical_rms": 0.06017461988580678, + "teaching_rms": 0.0013291754192877106, + "traffic_rms": 0.06015385317018739 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.8152776110346167e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5753545294706784e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021914178105991743, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011285493536293507 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031352996826171875, + "mean_feedback_forward_cosine": 0.9997590587985131, + "mean_feedback_forward_relative_error": 0.021585262170241724, + "min_feedback_forward_cosine": 0.9995101690292358 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014312083911910652, + "instruction_rms": 0.0014312083913766652, + "raw_apical_rms": 0.060158212981470914, + "teaching_rms": 0.0014312083911910652, + "traffic_rms": 0.06013395895160597 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.6009224995959812e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.578689555011155e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.022159492507651878, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.012068520529419183 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03134904429316521, + "mean_feedback_forward_cosine": 0.9997591280168102, + "mean_feedback_forward_relative_error": 0.021581231646480098, + "min_feedback_forward_cosine": 0.9995102286338806 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00128285456991729, + "instruction_rms": 0.0012828545704672103, + "raw_apical_rms": 0.060178350561723784, + "teaching_rms": 0.00128285456991729, + "traffic_rms": 0.060158784222565785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5508404277729824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5758511907512554e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021423602013939738, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.010624613016620278 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313451923429966, + "mean_feedback_forward_cosine": 0.9997592145396818, + "mean_feedback_forward_relative_error": 0.02157715622395758, + "min_feedback_forward_cosine": 0.9995103478431702 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013133976218261839, + "instruction_rms": 0.0013133976215214175, + "raw_apical_rms": 0.06014330045693746, + "teaching_rms": 0.0013133976218261839, + "traffic_rms": 0.060121834679379836 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3491713463386077e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5705432305124935e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02159263706327977, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.010737270039469003 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03134067729115486, + "mean_feedback_forward_cosine": 0.9997593222125885, + "mean_feedback_forward_relative_error": 0.021572456409734105, + "min_feedback_forward_cosine": 0.9995105266571045 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001391305149665331, + "instruction_rms": 0.0013913051489891754, + "raw_apical_rms": 0.060125212189627956, + "teaching_rms": 0.001391305149665331, + "traffic_rms": 0.06010070944313869 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9454923005923774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5803998891754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02168214473700549, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.011407286370545626 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03133620694279671, + "mean_feedback_forward_cosine": 0.9997594164263818, + "mean_feedback_forward_relative_error": 0.02156771686408789, + "min_feedback_forward_cosine": 0.9995107650756836 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013869202438614547, + "instruction_rms": 0.001386920243428907, + "raw_apical_rms": 0.06013829142432585, + "teaching_rms": 0.0013869202438614547, + "traffic_rms": 0.0601145159755107 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1140859729484873e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5793187245614857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021739876079040547, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.011116878349781036 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031331513077020645, + "mean_feedback_forward_cosine": 0.999759518331097, + "mean_feedback_forward_relative_error": 0.02156275722588743, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001417679202033186, + "instruction_rms": 0.0014176792018678064, + "raw_apical_rms": 0.06012870803794755, + "teaching_rms": 0.001417679202033186, + "traffic_rms": 0.06010324714483092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2602602456345267e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5714074246786214e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021847983224535093, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.012106659239511937 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132714703679085, + "mean_feedback_forward_cosine": 0.9997595971630465, + "mean_feedback_forward_relative_error": 0.021558043907486623, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001379679893926087, + "instruction_rms": 0.0013796798941189312, + "raw_apical_rms": 0.060092781516599685, + "teaching_rms": 0.001379679893926087, + "traffic_rms": 0.06006935939637903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.620488537274923e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570917304365128e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021682985947470337, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.010919339765757322 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132709860801697, + "mean_feedback_forward_cosine": 0.9997596010085075, + "mean_feedback_forward_relative_error": 0.021557943354691227, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001179968784308093, + "instruction_rms": 0.0011799687843855847, + "raw_apical_rms": 0.06008720547514168, + "teaching_rms": 0.001179968784308093, + "traffic_rms": 0.06006945434673334 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8576384874791074e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5612939849153663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021591201577678427, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.009476273037791251 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132707253098488, + "mean_feedback_forward_cosine": 0.9997596240812733, + "mean_feedback_forward_relative_error": 0.021557855254579938, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011296071960121053, + "instruction_rms": 0.0011296071964631573, + "raw_apical_rms": 0.06008207813474541, + "teaching_rms": 0.0011296071960121053, + "traffic_rms": 0.06006563827661422 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1674486205824905e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563976340381333e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021447013086679213, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.008430601634383202 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031327035278081894, + "mean_feedback_forward_cosine": 0.999759599085777, + "mean_feedback_forward_relative_error": 0.021557764721012885, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001121324080897597, + "instruction_rms": 0.0011213240807696613, + "raw_apical_rms": 0.06007834927889558, + "teaching_rms": 0.001121324080897597, + "traffic_rms": 0.06006248524135712 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2669075505632463e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5626452615872206e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021487707526781655, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008414895274192095 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031327009201049805, + "mean_feedback_forward_cosine": 0.9997596279267342, + "mean_feedback_forward_relative_error": 0.021557679745338617, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010462794952962285, + "instruction_rms": 0.001046279495115535, + "raw_apical_rms": 0.06007435809635876, + "teaching_rms": 0.0010462794952962285, + "traffic_rms": 0.06005984553676592 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9332576428610082e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5683413025204685e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02157790363457604, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.007825479854792356 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326986849308014, + "mean_feedback_forward_cosine": 0.9997595952403161, + "mean_feedback_forward_relative_error": 0.021557595625880263, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001036368860131206, + "instruction_rms": 0.0010363688600510232, + "raw_apical_rms": 0.06006759080480991, + "teaching_rms": 0.001036368860131206, + "traffic_rms": 0.06005341464384376 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9908087622866333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5634408615326202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021698631335063577, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.007752396997436881 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326957046985626, + "mean_feedback_forward_cosine": 0.9997596375403865, + "mean_feedback_forward_relative_error": 0.02155751351928038, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00103690814016864, + "instruction_rms": 0.001036908140151119, + "raw_apical_rms": 0.060066664923446415, + "teaching_rms": 0.00103690814016864, + "traffic_rms": 0.060052171413669905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.945481642451341e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5669167082605052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021475271499250284, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.007858595271185041 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326938420534134, + "mean_feedback_forward_cosine": 0.9997596183130818, + "mean_feedback_forward_relative_error": 0.021557435333248106, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009805716767374976, + "instruction_rms": 0.0009805716764035783, + "raw_apical_rms": 0.06006998751423511, + "teaching_rms": 0.0009805716767374976, + "traffic_rms": 0.060057039953904 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.433953127933819e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.56538217900001e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02124931950437315, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007328558262661099 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132690116763115, + "mean_feedback_forward_cosine": 0.999759599085777, + "mean_feedback_forward_relative_error": 0.021557354698738745, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010845778775543477, + "instruction_rms": 0.0010845778777478423, + "raw_apical_rms": 0.060069209873066264, + "teaching_rms": 0.0010845778775543477, + "traffic_rms": 0.06005369473041427 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9106012771885617e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.564227473703303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021559755123017183, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.007954697859212757 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132687509059906, + "mean_feedback_forward_cosine": 0.9997596260040037, + "mean_feedback_forward_relative_error": 0.02155727277239484, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010299564696549968, + "instruction_rms": 0.0010299564692812465, + "raw_apical_rms": 0.06006812192226086, + "teaching_rms": 0.0010299564696549968, + "traffic_rms": 0.060053931538063196 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.3349590705379342e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5660258580619945e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021729885976055806, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.007465883650183677 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326841562986374, + "mean_feedback_forward_cosine": 0.9997596144676208, + "mean_feedback_forward_relative_error": 0.021557185423350143, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001063413505157721, + "instruction_rms": 0.0010634135050796281, + "raw_apical_rms": 0.060071878818973384, + "teaching_rms": 0.001063413505157721, + "traffic_rms": 0.06005699924020882 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.3217483047233145e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5665946369244484e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021747135661781707, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007692031290978193 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132681921124458, + "mean_feedback_forward_cosine": 0.9997596298494646, + "mean_feedback_forward_relative_error": 0.02155710408283818, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010305865891414153, + "instruction_rms": 0.0010305865887556003, + "raw_apical_rms": 0.060076575512097774, + "teaching_rms": 0.0010305865891414153, + "traffic_rms": 0.06006234825377816 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.8926501255123185e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5669840817105678e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02140562166712032, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.00782879565000534 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313267819583416, + "mean_feedback_forward_cosine": 0.9997596086994294, + "mean_feedback_forward_relative_error": 0.021557018070692015, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001129956087231944, + "instruction_rms": 0.0011299560868735551, + "raw_apical_rms": 0.06007572241114544, + "teaching_rms": 0.001129956087231944, + "traffic_rms": 0.06005868079532163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.4017687394461973e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567557156080132e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021886043922274504, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.008396395660117269 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132675588130951, + "mean_feedback_forward_cosine": 0.9997596221585427, + "mean_feedback_forward_relative_error": 0.021556935483409513, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009887630050392931, + "instruction_rms": 0.0009887630052253591, + "raw_apical_rms": 0.06007528882168374, + "teaching_rms": 0.0009887630050392931, + "traffic_rms": 0.06006227369066905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.066620830021293e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5659746742414652e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021141643553731288, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0073443680226802826 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132672607898712, + "mean_feedback_forward_cosine": 0.9997596317721952, + "mean_feedback_forward_relative_error": 0.021556853451916286, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010581566164695499, + "instruction_rms": 0.0010581566167576506, + "raw_apical_rms": 0.060078922021673004, + "teaching_rms": 0.0010581566164695499, + "traffic_rms": 0.060063899841260364 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.0862328753423753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5685245696508257e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021514753858243214, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.007455915867015719 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326696276664734, + "mean_feedback_forward_cosine": 0.9997596375403865, + "mean_feedback_forward_relative_error": 0.021556774845286723, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010244626963122443, + "instruction_rms": 0.0010244626963263736, + "raw_apical_rms": 0.06008215650066717, + "teaching_rms": 0.0010244626963122443, + "traffic_rms": 0.060067852865176886 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.074700944376673e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.569071509815174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021306631787898232, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.0073924257943034175 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326670199632645, + "mean_feedback_forward_cosine": 0.9997596221585427, + "mean_feedback_forward_relative_error": 0.02155669374511607, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009767126207487105, + "instruction_rms": 0.0009767126210894945, + "raw_apical_rms": 0.06008434921583812, + "teaching_rms": 0.0009767126207487105, + "traffic_rms": 0.06007130250795151 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1931301219524357e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5672551918101414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021793107940057333, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0070690033512562514 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326647847890854, + "mean_feedback_forward_cosine": 0.9997596163903514, + "mean_feedback_forward_relative_error": 0.021556613831630638, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010006242303614047, + "instruction_rms": 0.0010006242309953588, + "raw_apical_rms": 0.06008041334988278, + "teaching_rms": 0.0010006242303614047, + "traffic_rms": 0.060066208644533915 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.7223707732559888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5673656243708937e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021532127395720854, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.007275210626721382 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326621770858765, + "mean_feedback_forward_cosine": 0.999759643308578, + "mean_feedback_forward_relative_error": 0.021556533617718567, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009555785630081935, + "instruction_rms": 0.0009555785630237475, + "raw_apical_rms": 0.0600744885089705, + "teaching_rms": 0.0009555785630081935, + "traffic_rms": 0.060062183466376456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 1.9061033640355163e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568538288493778e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021830055476936984, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006774568736106158 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132659196853638, + "mean_feedback_forward_cosine": 0.9997596375403865, + "mean_feedback_forward_relative_error": 0.021556453749297128, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009395257352998124, + "instruction_rms": 0.0009395257354966325, + "raw_apical_rms": 0.060068914411747804, + "teaching_rms": 0.0009395257352998124, + "traffic_rms": 0.0600566705271883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3887768207941917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5657385022341837e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021918999249193983, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006736315014958381 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132656216621399, + "mean_feedback_forward_cosine": 0.9997596336949256, + "mean_feedback_forward_relative_error": 0.021556375142667564, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009602019105545089, + "instruction_rms": 0.000960201910210807, + "raw_apical_rms": 0.06006304963139241, + "teaching_rms": 0.0009602019105545089, + "traffic_rms": 0.06005012476332571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.4388381092421696e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5667665494101504e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02127287783141234, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.007116384837180376 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313265360891819, + "mean_feedback_forward_cosine": 0.9997596317721952, + "mean_feedback_forward_relative_error": 0.021556297782808542, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009628185131176399, + "instruction_rms": 0.0009628185128434622, + "raw_apical_rms": 0.060057812696817314, + "teaching_rms": 0.0009628185131176399, + "traffic_rms": 0.060044823555564404 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.9614679658275236e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5639761409025163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02162741097611953, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.007078839312195778 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132651373744011, + "mean_feedback_forward_cosine": 0.999759643308578, + "mean_feedback_forward_relative_error": 0.02155622180491205, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009328463536858388, + "instruction_rms": 0.0009328463533456953, + "raw_apical_rms": 0.060054184034077036, + "teaching_rms": 0.0009328463536858388, + "traffic_rms": 0.06004251195105212 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.343368521451339e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568146097608532e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02127933575466516, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0068743041935563085 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132648766040802, + "mean_feedback_forward_cosine": 0.9997596356176561, + "mean_feedback_forward_relative_error": 0.02155614274764253, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009752987427118393, + "instruction_rms": 0.0009752987428624305, + "raw_apical_rms": 0.06005889848482817, + "teaching_rms": 0.0009752987427118393, + "traffic_rms": 0.060045733577735357 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.312624758360471e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5669000475049028e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021269330889308057, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006992063255682588 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132646530866623, + "mean_feedback_forward_cosine": 0.9997596240812733, + "mean_feedback_forward_relative_error": 0.021556065102378206, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009754219255835325, + "instruction_rms": 0.0009754219253127989, + "raw_apical_rms": 0.060058905286350696, + "teaching_rms": 0.0009754219255835325, + "traffic_rms": 0.06004608744551165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0104115705521508e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5657628319952628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021313992081019063, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.007167713696211576 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132643550634384, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.0215559869463886, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009615485037101152, + "instruction_rms": 0.000961548503591037, + "raw_apical_rms": 0.060060380172654845, + "teaching_rms": 0.0009615485037101152, + "traffic_rms": 0.06004752553417286 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.7993767304224093e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5648675156693924e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021576789954253533, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006852393444180488 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326405704021454, + "mean_feedback_forward_cosine": 0.9997596336949256, + "mean_feedback_forward_relative_error": 0.021555908159503052, + "min_feedback_forward_cosine": 0.9995108246803284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009785713771199437, + "instruction_rms": 0.0009785713774130107, + "raw_apical_rms": 0.060061160174462176, + "teaching_rms": 0.0009785713771199437, + "traffic_rms": 0.060047543608914046 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.949755912278306e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5666668773173498e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021418184265572616, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006754859584569931 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132638707756996, + "mean_feedback_forward_cosine": 0.9997596317721952, + "mean_feedback_forward_relative_error": 0.02155583209147857, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009655848167913925, + "instruction_rms": 0.0009655848170674031, + "raw_apical_rms": 0.060062541633444995, + "teaching_rms": 0.0009655848167913925, + "traffic_rms": 0.06004947472969154 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0814971080085343e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567244652274971e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021253830157798064, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.006898634309321642 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132636100053787, + "mean_feedback_forward_cosine": 0.9997596375403865, + "mean_feedback_forward_relative_error": 0.021555756954776663, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009420714872473244, + "instruction_rms": 0.0009420714869762677, + "raw_apical_rms": 0.060060731586379354, + "teaching_rms": 0.0009420714872473244, + "traffic_rms": 0.0600480769955571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.1064197497366877e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5663262435928885e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021561401438528468, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.006688732709065079 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132633492350578, + "mean_feedback_forward_cosine": 0.9997596317721952, + "mean_feedback_forward_relative_error": 0.02155567592971267, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010051458416884568, + "instruction_rms": 0.0010051458409989927, + "raw_apical_rms": 0.06006076079226069, + "teaching_rms": 0.0010051458416884568, + "traffic_rms": 0.060046603722024645 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1515237591529512e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5718607030157967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02122170693276716, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.007178794793337584 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326305121183395, + "mean_feedback_forward_cosine": 0.9997596163903514, + "mean_feedback_forward_relative_error": 0.02155560139386404, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009334209579252473, + "instruction_rms": 0.0009334209576963297, + "raw_apical_rms": 0.06005812042541105, + "teaching_rms": 0.0009334209579252473, + "traffic_rms": 0.060045622230531406 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5771383249661994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670988070826762e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021125596140977525, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006375368404537439 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326279044151306, + "mean_feedback_forward_cosine": 0.9997596413858475, + "mean_feedback_forward_relative_error": 0.02155552347821574, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009267334638879396, + "instruction_rms": 0.0009267334636475204, + "raw_apical_rms": 0.060054398744299894, + "teaching_rms": 0.0009267334638879396, + "traffic_rms": 0.060042373605508195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.162148327045088e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.564611431806649e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02162797211756293, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006588259144797921 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326256692409515, + "mean_feedback_forward_cosine": 0.9997596221585427, + "mean_feedback_forward_relative_error": 0.021555449888711015, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008881864527277518, + "instruction_rms": 0.0008881864527954909, + "raw_apical_rms": 0.06004703374881644, + "teaching_rms": 0.0008881864527277518, + "traffic_rms": 0.06003612336996302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.7895343873419733e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568090316629994e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021218560867508986, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.00602709147669375 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132622316479683, + "mean_feedback_forward_cosine": 0.999759647154039, + "mean_feedback_forward_relative_error": 0.02155537859747006, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009482835371602856, + "instruction_rms": 0.0009482835371945668, + "raw_apical_rms": 0.060052149373275965, + "teaching_rms": 0.0009482835371602856, + "traffic_rms": 0.060039407708248935 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3784711089547272e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5660439395201268e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021661237161664414, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006757023540996015 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132620081305504, + "mean_feedback_forward_cosine": 0.9997596394631171, + "mean_feedback_forward_relative_error": 0.021555301538037676, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000881289653620672, + "instruction_rms": 0.0008812896537584988, + "raw_apical_rms": 0.06004567477508321, + "teaching_rms": 0.000881289653620672, + "traffic_rms": 0.06003469592632789 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.111737984478168e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5672102214083974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020971044230362227, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0061257558731734755 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326182186603546, + "mean_feedback_forward_cosine": 0.9997596298494646, + "mean_feedback_forward_relative_error": 0.0215552277832983, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009306139623574617, + "instruction_rms": 0.0009306139624281346, + "raw_apical_rms": 0.06004385099800795, + "teaching_rms": 0.0009306139623574617, + "traffic_rms": 0.060030674162966 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.143248956087973e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568642914579012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02127893229936149, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0065580026246607305 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132615238428116, + "mean_feedback_forward_cosine": 0.9997596298494646, + "mean_feedback_forward_relative_error": 0.021555148726028782, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009033962134555074, + "instruction_rms": 0.0009033962133567819, + "raw_apical_rms": 0.06004334921704471, + "teaching_rms": 0.0009033962134555074, + "traffic_rms": 0.06003160378592111 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.932926707581828e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5649429756040186e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0212443712795138, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006223944382146001 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132612258195877, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.021555073844689514, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009210895115788966, + "instruction_rms": 0.0009210895113247055, + "raw_apical_rms": 0.06004048408063149, + "teaching_rms": 0.0009210895115788966, + "traffic_rms": 0.06002838228205816 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.5178549023507912e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5697126001669134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021230850876854702, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006426349698528648 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132610395550728, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.021555001111400705, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009018071280486433, + "instruction_rms": 0.0009018071281110626, + "raw_apical_rms": 0.060041556149045, + "teaching_rms": 0.0009018071280486433, + "traffic_rms": 0.060030215267169465 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.169271517971083e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5668077305435055e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021291286597336086, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.006350258146673441 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132607787847519, + "mean_feedback_forward_cosine": 0.9997596356176561, + "mean_feedback_forward_relative_error": 0.02155492882875185, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008698168059557551, + "instruction_rms": 0.0008698168062776598, + "raw_apical_rms": 0.060039656360668346, + "teaching_rms": 0.0008698168059557551, + "traffic_rms": 0.060029301023976 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3856882691575265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5680749689917897e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021268179344269673, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006042736135423183 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326062977313995, + "mean_feedback_forward_cosine": 0.9997596529222303, + "mean_feedback_forward_relative_error": 0.0215548605117346, + "min_feedback_forward_cosine": 0.9995111227035522 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000836438271041538, + "instruction_rms": 0.0008364382711784785, + "raw_apical_rms": 0.060043254627502034, + "teaching_rms": 0.000836438271041538, + "traffic_rms": 0.06003299805436287 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.36579200674214e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5717391871834188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021861043911415013, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0059654509999603035 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0313260443508625, + "mean_feedback_forward_cosine": 0.9997596548449609, + "mean_feedback_forward_relative_error": 0.021554786426525925, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008727192140346393, + "instruction_rms": 0.0008727192142567254, + "raw_apical_rms": 0.06003745986382582, + "teaching_rms": 0.0008727192140346393, + "traffic_rms": 0.06002682889271375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.894366796761915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5717489389259885e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02137953377467197, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006041391937145963 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031326014548540115, + "mean_feedback_forward_cosine": 0.9997596606131522, + "mean_feedback_forward_relative_error": 0.02155471441426104, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008935821916719584, + "instruction_rms": 0.0008935821915541255, + "raw_apical_rms": 0.060039934075795, + "teaching_rms": 0.0008935821916719584, + "traffic_rms": 0.06002910980464427 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.1903458602423598e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5688957457962854e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021560479270196505, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.006232946597337723 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031325992196798325, + "mean_feedback_forward_cosine": 0.9997596490767694, + "mean_feedback_forward_relative_error": 0.021554640043647058, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009259972258659079, + "instruction_rms": 0.0009259972258051484, + "raw_apical_rms": 0.060043990302327876, + "teaching_rms": 0.0009259972258659079, + "traffic_rms": 0.06003191105315353 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.2353193074309274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5704890834058506e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021402499272677757, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.006447158992290497 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132596239447594, + "mean_feedback_forward_cosine": 0.9997596740722656, + "mean_feedback_forward_relative_error": 0.021554567039974274, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009317201922527884, + "instruction_rms": 0.0009317201921949359, + "raw_apical_rms": 0.060040267249132, + "teaching_rms": 0.0009317201922527884, + "traffic_rms": 0.06002745013319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.1710826914045356e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5729904493764114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021714220810872112, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.006646980301588773 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132593631744385, + "mean_feedback_forward_cosine": 0.9997596356176561, + "mean_feedback_forward_relative_error": 0.02155449348051221, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009172148378231025, + "instruction_rms": 0.0009172148380027723, + "raw_apical_rms": 0.060044784582668285, + "teaching_rms": 0.0009172148378231025, + "traffic_rms": 0.06003272725118342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.132829202139419e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5691497142614578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021261427735506933, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006242364752627909 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132592514157295, + "mean_feedback_forward_cosine": 0.9997596317721952, + "mean_feedback_forward_relative_error": 0.021554420792287397, + "min_feedback_forward_cosine": 0.9995108842849731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008394010266712181, + "instruction_rms": 0.0008394010263244715, + "raw_apical_rms": 0.06003570661043683, + "teaching_rms": 0.0008394010266712181, + "traffic_rms": 0.060025402588168804 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 1.893620549253683e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5693063478141967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02136412229458273, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.005690691779181361 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132590278983116, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.02155434744312398, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000894822445704649, + "instruction_rms": 0.0008948224458492012, + "raw_apical_rms": 0.06003452088921942, + "teaching_rms": 0.000894822445704649, + "traffic_rms": 0.06002304114955634 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.3364913559476008e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.572832006419398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02151670856561304, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.00621510066986084 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.031325872987508774, + "mean_feedback_forward_cosine": 0.999759647154039, + "mean_feedback_forward_relative_error": 0.021554270879395546, + "min_feedback_forward_cosine": 0.9995110034942627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008731963482363144, + "instruction_rms": 0.0008731963485391498, + "raw_apical_rms": 0.06003501312016347, + "teaching_rms": 0.0008731963482363144, + "traffic_rms": 0.06002380693724977 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.2314685210744756e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5687875862690624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02158584079171715, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.006227425732985139 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132585063576698, + "mean_feedback_forward_cosine": 0.9997596375403865, + "mean_feedback_forward_relative_error": 0.021554195201925693, + "min_feedback_forward_cosine": 0.9995110630989075 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008942622408165716, + "instruction_rms": 0.0008942622404361476, + "raw_apical_rms": 0.060033859223595505, + "teaching_rms": 0.0008942622408165716, + "traffic_rms": 0.06002265608609162 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146561145782471, + "maximum_absolute_post_projection_soma_slope": 2.0849823201274376e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670899861723797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.021497920676601043, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.006085729780197144 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03132582828402519, + "mean_feedback_forward_cosine": 0.9997596452313084, + "mean_feedback_forward_relative_error": 0.021554123805535418, + "min_feedback_forward_cosine": 0.9995109438896179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008740050133278148, + "instruction_rms": 0.0008740050134007767, + "raw_apical_rms": 0.06003476288060316, + "teaching_rms": 0.0008740050133278148, + "traffic_rms": 0.06002387793777457 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.07146560400724411, + "maximum_absolute_post_projection_soma_slope": 2.2439492042281017e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5722066097305383e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0217050617177861, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0059590229891985655 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9217, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.387104931640625, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2529336320, + "peak_memory_reserved_bytes": 3821010944, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 3.0876691425874014e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 3.06026421981187e-08, + "min_residual_soma_slope": -3.0876691425874014e-08, + "mse": 3.3290338960027444e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.160152391392168e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.160152391392168e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1395759582519531, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.040845394134521484, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 14820.786049365997, + "train_wall_s": 14817.008256196976 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.016692832112312317, + 0.01200532540678978, + 0.008332769386470318, + 0.0075498903170228004, + 0.0053708222694695, + 0.006087752059102058, + 0.003867587074637413, + 0.005006805062294006, + 0.002887815935537219, + 0.003748569404706359, + 0.0022844804916530848, + 0.004654052667319775, + 0.0032952053006738424, + 0.004196343943476677, + 0.002609694842249155, + 0.0035618862602859735, + 0.0021939175203442574, + 0.0029503554105758667, + 0.0018851794302463531, + 0.0024613135028630495, + 0.0017383317463099957, + 0.0038990096654742956, + 0.0027804728597402573, + 0.003515329910442233, + 0.0024628741666674614, + 0.003224294865503907, + 0.0022232758346945047, + 0.002873069839552045, + 0.0020843856036663055, + 0.0026347259990870953, + 0.0020947945304214954 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.000000476837158, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.08893077075481415, + 0.062396932393312454, + 0.029449449852108955, + 0.03909926116466522, + 0.014590166509151459, + 0.03232475742697716, + 0.008942865766584873, + 0.02675946056842804, + 0.00599529268220067, + 0.019410571083426476, + 0.004170955624431372, + 0.02429185062646866, + 0.00736309215426445, + 0.02235197275876999, + 0.005150705110281706, + 0.01915901154279709, + 0.0039020790718495846, + 0.015739649534225464, + 0.0030583241023123264, + 0.01274823397397995, + 0.00259096990339458, + 0.02055993489921093, + 0.005347526632249355, + 0.018513329327106476, + 0.0042116641998291016, + 0.017480242997407913, + 0.0035575039219111204, + 0.015222329646348953, + 0.003111188765615225, + 0.013976861722767353, + 0.0029232106171548367 + ], + "unscaled_traffic_rms": [ + 0.7508236765861511, + 0.7696099877357483, + 1.131806492805481, + 0.7723819017410278, + 1.4724498987197876, + 0.7533237934112549, + 1.7299095392227173, + 0.7484164237976074, + 1.9267221689224243, + 0.7724800109863281, + 2.1908462047576904, + 0.7663562297821045, + 1.7901203632354736, + 0.7509572505950928, + 2.026669979095459, + 0.7436471581459045, + 2.2489728927612305, + 0.7497893571853638, + 2.46563720703125, + 0.7722837328910828, + 2.6836771965026855, + 0.7585645914077759, + 2.079819679260254, + 0.7595241069793701, + 2.3390982151031494, + 0.7378146648406982, + 2.4998154640197754, + 0.7549619078636169, + 2.6798574924468994, + 0.7540250420570374, + 2.8664298057556152 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684206618894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 72745580756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 2745665106100224 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json new file mode 100644 index 0000000..c7e1fea --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json @@ -0,0 +1,6861 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5014, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9996523022651672, + "feedback_forward_cosine": [ + 0.9998147487640381, + 0.9998278617858887, + 0.9998444318771362, + 0.999819278717041, + 0.9996131658554077, + 0.9994606971740723, + 0.9997701644897461, + 0.9996477365493774, + 0.9996994733810425, + 0.9995297193527222, + 0.9996874928474426, + 0.9997687339782715, + 0.9997282028198242, + 0.9996908903121948, + 0.9997670650482178, + 0.9997155666351318, + 0.9997864961624146, + 0.9997143745422363, + 0.9998044967651367, + 0.9997119307518005, + 0.9997832179069519, + 0.9998683929443359, + 0.9998384714126587, + 0.999819278717041, + 0.9998373985290527, + 0.9997920989990234, + 0.9998111128807068, + 0.9997344017028809, + 0.9998141527175903, + 0.9995765686035156, + 0.9999843835830688 + ], + "feedback_forward_norm_ratio": [ + 0.9988206624984741, + 1.001134991645813, + 0.9998640418052673, + 1.0007452964782715, + 0.9991434812545776, + 1.0023818016052246, + 0.9996303915977478, + 1.0020372867584229, + 0.9989871382713318, + 1.0012645721435547, + 0.9983092546463013, + 1.0002796649932861, + 0.9997096657752991, + 1.0010067224502563, + 0.9992865920066833, + 1.001634120941162, + 0.9995214343070984, + 1.0006155967712402, + 0.9994485974311829, + 1.0013827085494995, + 0.9992908239364624, + 1.0003712177276611, + 0.9996266961097717, + 1.0006937980651855, + 0.99983149766922, + 1.000633955001831, + 0.9995362758636475, + 1.0009021759033203, + 0.9998878240585327, + 1.001349687576294, + 1.0026386976242065 + ], + "feedback_forward_relative_error": [ + 0.019273990765213966, + 0.018598997965455055, + 0.017638105899095535, + 0.01903495378792286, + 0.0278138630092144, + 0.03296716883778572, + 0.02143999934196472, + 0.026652244850993156, + 0.024521373212337494, + 0.03071632608771324, + 0.02503705956041813, + 0.021511593833565712, + 0.023311004042625427, + 0.024900957942008972, + 0.02158486656844616, + 0.02392916940152645, + 0.02066095359623432, + 0.02391694486141205, + 0.019776079803705215, + 0.024060113355517387, + 0.020828573033213615, + 0.016229894012212753, + 0.017972342669963837, + 0.019035600125789642, + 0.018028639256954193, + 0.020411977544426918, + 0.01943998970091343, + 0.023075181990861893, + 0.019283244386315346, + 0.029152056202292442, + 0.006188182160258293 + ], + "innovation_negative_gradient_cosine": [ + 0.9996793270111084, + 0.9996604919433594, + 0.9996795654296875, + 0.9996747970581055, + 0.9996909499168396, + 0.9995384216308594, + 0.9996996521949768, + 0.9996235370635986, + 0.9997107982635498, + 0.9995654821395874, + 0.9997165203094482, + 0.9996768236160278, + 0.9997236132621765, + 0.999600887298584, + 0.999730110168457, + 0.9996433258056641, + 0.9997373819351196, + 0.9996926784515381, + 0.9997440576553345, + 0.9996826648712158, + 0.9997591376304626, + 0.9997883439064026, + 0.9997954964637756, + 0.9997599720954895, + 0.9998281598091125, + 0.9998058676719666, + 0.9998501539230347, + 0.9998639822006226, + 0.99992835521698, + 0.9999697208404541, + 0.8102250099182129 + ], + "instruction_negative_gradient_cosine": [ + 0.9996793866157532, + 0.9996604919433594, + 0.9996795654296875, + 0.9996747970581055, + 0.9996908903121948, + 0.9995384216308594, + 0.9996996521949768, + 0.9996236562728882, + 0.9997107982635498, + 0.9995654821395874, + 0.9997165203094482, + 0.9996768236160278, + 0.9997236132621765, + 0.999600887298584, + 0.999730110168457, + 0.9996433258056641, + 0.9997373819351196, + 0.9996926784515381, + 0.9997440576553345, + 0.9996826648712158, + 0.9997591376304626, + 0.9997883439064026, + 0.9997954964637756, + 0.9997599720954895, + 0.9998281002044678, + 0.9998058080673218, + 0.9998501539230347, + 0.9998640418052673, + 0.99992835521698, + 0.9999697804450989, + 0.8450019359588623 + ], + "matched_negative_gradient_cosine": [ + 0.019015047699213028, + -8.82298918440938e-05, + 0.0423874706029892, + -0.02354944497346878, + 0.06672713160514832, + 0.03335469216108322, + 0.05295939743518829, + 0.09118880331516266, + 0.039782822132110596, + -0.001272554975003004, + 0.04149865731596947, + 0.02153804898262024, + 0.013430852442979813, + -0.001538371667265892, + 0.016360076144337654, + 0.029660258442163467, + 0.02270975336432457, + 0.0486227348446846, + 0.02294677123427391, + 0.045042991638183594, + 0.013027578592300415, + 0.00554568599909544, + -0.054277826100587845, + 0.08001662790775299, + -0.019213473424315453, + 0.1141226664185524, + -0.0042828223668038845, + 0.19512201845645905, + 0.012721724808216095, + 0.010696874000132084, + 0.11020606756210327 + ], + "max_norm_match_direction_error": 0.006641268730163574, + "max_norm_match_relative_error": 1.5802008590526384e-07, + "mean_feedback_forward_cosine": 0.9997503872840635, + "mean_feedback_forward_relative_error": 0.021838433800205108, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.10311372578144073, + "max_absolute_post_projection_soma_slope": 2.0606622186392087e-08, + "max_absolute_pre_projection_soma_slope": 0.10311372578144073, + "max_positive_post_projection_soma_slope": 2.0606622186392087e-08, + "min_post_projection_soma_slope": -1.060392040841407e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.3967742735909173e-08, + "pre_projection_traffic_rms_ratio": 0.008809628894839435 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.008809628894815251, + "raw_negative_gradient_cosine": [ + 0.019015047699213028, + -8.822965901345015e-05, + 0.0423874706029892, + -0.02354944497346878, + 0.06672713160514832, + 0.03335469216108322, + 0.05295940116047859, + 0.09118880331516266, + 0.039782822132110596, + -0.001272555673494935, + 0.04149865731596947, + 0.02153804898262024, + 0.013430852442979813, + -0.001538372365757823, + 0.016360074281692505, + 0.029660258442163467, + 0.02270975150167942, + 0.0486227348446846, + 0.02294677123427391, + 0.045042991638183594, + 0.01302757766097784, + 0.0055456869304180145, + -0.05427782982587814, + 0.08001662790775299, + -0.0192134790122509, + 0.114122673869133, + -0.004282823298126459, + 0.19512201845645905, + 0.012721724808216095, + 0.010696874931454659, + 0.11020593345165253 + ], + "teaching_negative_gradient_cosine": [ + 0.9996793270111084, + 0.9996604919433594, + 0.9996795654296875, + 0.9996747970581055, + 0.9996909499168396, + 0.9995384216308594, + 0.9996996521949768, + 0.9996235370635986, + 0.9997107982635498, + 0.9995654821395874, + 0.9997165203094482, + 0.9996768236160278, + 0.9997236132621765, + 0.999600887298584, + 0.999730110168457, + 0.9996433258056641, + 0.9997373819351196, + 0.9996926784515381, + 0.9997440576553345, + 0.9996826648712158, + 0.9997591376304626, + 0.9997883439064026, + 0.9997954964637756, + 0.9997599720954895, + 0.9998281598091125, + 0.9998058676719666, + 0.9998501539230347, + 0.9998639822006226, + 0.99992835521698, + 0.9999697208404541, + 0.8102250099182129 + ], + "traffic_instruction_rms_ratio": 80.5082313124881, + "used_negative_gradient_cosine": [ + 0.9996793270111084, + 0.9996604919433594, + 0.9996795654296875, + 0.9996747970581055, + 0.9996909499168396, + 0.9995384216308594, + 0.9996996521949768, + 0.9996235370635986, + 0.9997107982635498, + 0.9995654821395874, + 0.9997165203094482, + 0.9996768236160278, + 0.9997236132621765, + 0.999600887298584, + 0.999730110168457, + 0.9996433258056641, + 0.9997373819351196, + 0.9996926784515381, + 0.9997440576553345, + 0.9996826648712158, + 0.9997591376304626, + 0.9997883439064026, + 0.9997954964637756, + 0.9997599720954895, + 0.9998281598091125, + 0.9998058676719666, + 0.9998501539230347, + 0.9998639822006226, + 0.99992835521698, + 0.9999697208404541, + 0.8102250099182129 + ], + "wall_s": 0.23915576934814453 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.349123239517212, + "mean_feedback_forward_cosine": 0.3184154016836997, + "mean_feedback_forward_relative_error": 1.1622308550342437, + "min_feedback_forward_cosine": 0.09792320430278778 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0018683632463321101, + "instruction_rms": 0.0018683632464356298, + "raw_apical_rms": 0.03624569453386514, + "teaching_rms": 0.0018683632463321101, + "traffic_rms": 0.03613115014401499 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 4.49112000922014e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.052199008723502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018145315137623224, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 1.9720316844940184 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3080101013183594, + "mean_feedback_forward_cosine": 0.38682587685123565, + "mean_feedback_forward_relative_error": 1.1022206256466527, + "min_feedback_forward_cosine": 0.14350903034210205 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002286403119993488, + "instruction_rms": 0.0022864031199263044, + "raw_apical_rms": 0.0415839044188337, + "teaching_rms": 0.002286403119993488, + "traffic_rms": 0.041520381929141176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 4.69071181896652e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7511637960013122e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018682236871780205, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.4413672492218017 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2625365257263184, + "mean_feedback_forward_cosine": 0.454744745166071, + "mean_feedback_forward_relative_error": 1.0403635992157845, + "min_feedback_forward_cosine": 0.2042173445224762 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002856331148956587, + "instruction_rms": 0.0028563311487369188, + "raw_apical_rms": 0.04831714058676078, + "teaching_rms": 0.002856331148956587, + "traffic_rms": 0.04823201347851221 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 4.657456642576108e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6100743518043677e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016788666455893747, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.1883209758377076 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2150771617889404, + "mean_feedback_forward_cosine": 0.5136532504712382, + "mean_feedback_forward_relative_error": 0.9834162862070145, + "min_feedback_forward_cosine": 0.26392054557800293 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031697569633185796, + "instruction_rms": 0.003169756963222006, + "raw_apical_rms": 0.05338747022468913, + "teaching_rms": 0.0031697569633185796, + "traffic_rms": 0.05329255701858028 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 5.36680921925381e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6000444448801e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015669059732735403, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 0.9779595841789246 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1636258363723755, + "mean_feedback_forward_cosine": 0.5632070437554391, + "mean_feedback_forward_relative_error": 0.9331383128320018, + "min_feedback_forward_cosine": 0.3270026445388794 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003303326608128507, + "instruction_rms": 0.00330332660864547, + "raw_apical_rms": 0.05613209391521324, + "teaching_rms": 0.003303326608128507, + "traffic_rms": 0.056034267856350774 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 6.213216607875438e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5812340690954248e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014205358990345121, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.8270183102226257 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1149184703826904, + "mean_feedback_forward_cosine": 0.6046481709326467, + "mean_feedback_forward_relative_error": 0.8884337246417999, + "min_feedback_forward_cosine": 0.3852089047431946 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033687365254883306, + "instruction_rms": 0.0033687365255427463, + "raw_apical_rms": 0.05853834170146622, + "teaching_rms": 0.0033687365254883306, + "traffic_rms": 0.058440623063614965 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.317833450315447e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5743119560269975e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013208046553298632, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.7185641108131409 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0763901472091675, + "mean_feedback_forward_cosine": 0.640215261328605, + "mean_feedback_forward_relative_error": 0.8478996340305575, + "min_feedback_forward_cosine": 0.4314519166946411 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003439691611151573, + "instruction_rms": 0.0034396916117434228, + "raw_apical_rms": 0.06018331462216512, + "teaching_rms": 0.003439691611151573, + "traffic_rms": 0.060083920337087586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.937631352206154e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5726156926374277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013604340684405218, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.6477009527969361 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0397318601608276, + "mean_feedback_forward_cosine": 0.6711982613609683, + "mean_feedback_forward_relative_error": 0.8106753114731081, + "min_feedback_forward_cosine": 0.47156092524528503 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034773293972525346, + "instruction_rms": 0.0034773293976698605, + "raw_apical_rms": 0.061118584269800776, + "teaching_rms": 0.0034773293972525346, + "traffic_rms": 0.061018724903928755 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 9.316102023149142e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5896862042897907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01365289767855389, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.5930870194435119 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0042730569839478, + "mean_feedback_forward_cosine": 0.6981356220860635, + "mean_feedback_forward_relative_error": 0.7769870152396541, + "min_feedback_forward_cosine": 0.5083057284355164 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034731553950747888, + "instruction_rms": 0.0034731553952689577, + "raw_apical_rms": 0.0626269252739908, + "teaching_rms": 0.0034731553950747888, + "traffic_rms": 0.06252966378164276 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0759437429896934e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.578908123350755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014179395672452142, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.548681717376709 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9682550430297852, + "mean_feedback_forward_cosine": 0.7225830458825634, + "mean_feedback_forward_relative_error": 0.7448044092424454, + "min_feedback_forward_cosine": 0.544492244720459 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003548925307316809, + "instruction_rms": 0.0035489253083595103, + "raw_apical_rms": 0.0627202408191766, + "teaching_rms": 0.003548925307316809, + "traffic_rms": 0.06261872837906626 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.151621161938237e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5809551235128574e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013916157228203477, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5209420721435547 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9336236715316772, + "mean_feedback_forward_cosine": 0.7443536039321653, + "mean_feedback_forward_relative_error": 0.7148607430919525, + "min_feedback_forward_cosine": 0.5774005651473999 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035479097876059545, + "instruction_rms": 0.0035479097873360125, + "raw_apical_rms": 0.06351287187237872, + "teaching_rms": 0.0035479097876059545, + "traffic_rms": 0.0634126927515392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 6.495565685327165e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5814357619928713e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014521128082131329, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.4921373353767395 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.899095892906189, + "mean_feedback_forward_cosine": 0.764203384999306, + "mean_feedback_forward_relative_error": 0.686450574667223, + "min_feedback_forward_cosine": 0.6088297963142395 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003567407650267457, + "instruction_rms": 0.003567407649679948, + "raw_apical_rms": 0.0643272545903745, + "teaching_rms": 0.003567407650267457, + "traffic_rms": 0.06422709151095458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0085728519015902e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5812116644453117e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014100058092938588, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.4646897285270691 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.869840681552887, + "mean_feedback_forward_cosine": 0.7823158867897526, + "mean_feedback_forward_relative_error": 0.6592727053549982, + "min_feedback_forward_cosine": 0.6380674839019775 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035551361837908983, + "instruction_rms": 0.003555136183908044, + "raw_apical_rms": 0.06526339862688703, + "teaching_rms": 0.0035551361837908983, + "traffic_rms": 0.06516528307975758 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1249581177708023e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5811661225743245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013975131741446984, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.44938543108940127 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8421918749809265, + "mean_feedback_forward_cosine": 0.7986098816317897, + "mean_feedback_forward_relative_error": 0.6339030655161026, + "min_feedback_forward_cosine": 0.6636968851089478 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003542310093047334, + "instruction_rms": 0.0035423100928636923, + "raw_apical_rms": 0.06566515295494509, + "teaching_rms": 0.003542310093047334, + "traffic_rms": 0.06556815349269349 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.069285673338527e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5747109182637625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014421003884813169, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4314415920257568 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8139985203742981, + "mean_feedback_forward_cosine": 0.8136870899508076, + "mean_feedback_forward_relative_error": 0.6094659340958442, + "min_feedback_forward_cosine": 0.6859352588653564 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00356102866105206, + "instruction_rms": 0.003561028660416645, + "raw_apical_rms": 0.06579666079935316, + "teaching_rms": 0.00356102866105206, + "traffic_rms": 0.06569893122453155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.28980972755744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5797627522924986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015149934872542212, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4118518581676483 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7864567637443542, + "mean_feedback_forward_cosine": 0.8278130292892456, + "mean_feedback_forward_relative_error": 0.5858500104758048, + "min_feedback_forward_cosine": 0.706699013710022 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003571298423541089, + "instruction_rms": 0.0035712984229070567, + "raw_apical_rms": 0.06652252201477747, + "teaching_rms": 0.003571298423541089, + "traffic_rms": 0.06642525712075328 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 7.55757980641647e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5685017447848382e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014987340263861476, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.39782868075370786 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7594925165176392, + "mean_feedback_forward_cosine": 0.8409136937510583, + "mean_feedback_forward_relative_error": 0.5628470672715095, + "min_feedback_forward_cosine": 0.7266198992729187 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036124114347874862, + "instruction_rms": 0.0036124114343832375, + "raw_apical_rms": 0.0667853896376609, + "teaching_rms": 0.0036124114347874862, + "traffic_rms": 0.06668641115497063 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.381794375760364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5778748470378126e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01494092720952503, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.3853364744567871 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7328860759735107, + "mean_feedback_forward_cosine": 0.8527568886356969, + "mean_feedback_forward_relative_error": 0.5412813420257261, + "min_feedback_forward_cosine": 0.7453172206878662 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036054019142679953, + "instruction_rms": 0.0036054019139316923, + "raw_apical_rms": 0.0669885631759906, + "teaching_rms": 0.0036054019142679953, + "traffic_rms": 0.06688996731069322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 8.959162300925527e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.574728810468514e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014829189195670446, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.37180454840660093 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7077186703681946, + "mean_feedback_forward_cosine": 0.8637468526440282, + "mean_feedback_forward_relative_error": 0.5204056270660893, + "min_feedback_forward_cosine": 0.7624452114105225 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003605624887831031, + "instruction_rms": 0.0036056248874183766, + "raw_apical_rms": 0.06821762426263715, + "teaching_rms": 0.003605624887831031, + "traffic_rms": 0.06812068622192712 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.481707342549271e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.56681578902328e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015647824216503382, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3612572445106506 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.683469831943512, + "mean_feedback_forward_cosine": 0.8740252410211871, + "mean_feedback_forward_relative_error": 0.5002171757721132, + "min_feedback_forward_cosine": 0.7783001661300659 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003619591928771839, + "instruction_rms": 0.0036195919276024928, + "raw_apical_rms": 0.06828244077412947, + "teaching_rms": 0.003619591928771839, + "traffic_rms": 0.06818475517578435 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.020572923011059e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5738955629962453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015268541183042194, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.348959853515625 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6594583988189697, + "mean_feedback_forward_cosine": 0.8832904523418795, + "mean_feedback_forward_relative_error": 0.48108004658452924, + "min_feedback_forward_cosine": 0.7932708263397217 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035751354082638317, + "instruction_rms": 0.0035751354080015697, + "raw_apical_rms": 0.06857471519851022, + "teaching_rms": 0.0035751354082638317, + "traffic_rms": 0.06848007569623833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 5.069496467058343e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5704569681491547e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015717778113856357, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3425276482486725 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6360369920730591, + "mean_feedback_forward_cosine": 0.891940749460651, + "mean_feedback_forward_relative_error": 0.46262357600273624, + "min_feedback_forward_cosine": 0.807397723197937 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036275424698628857, + "instruction_rms": 0.0036275424707179647, + "raw_apical_rms": 0.06835233988520548, + "teaching_rms": 0.0036275424698628857, + "traffic_rms": 0.0682543939807236 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 5.109001577352501e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5694009383727903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015192259002773798, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.32964239320755007 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6133252382278442, + "mean_feedback_forward_cosine": 0.9001314582363251, + "mean_feedback_forward_relative_error": 0.4446873885969962, + "min_feedback_forward_cosine": 0.8209010362625122 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003604449982899816, + "instruction_rms": 0.00360444998338726, + "raw_apical_rms": 0.06863894432059107, + "teaching_rms": 0.003604449982899816, + "traffic_rms": 0.06854259212185834 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 5.521335211255973e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5750633355662413e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015290812355600071, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3240155930995941 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5919969081878662, + "mean_feedback_forward_cosine": 0.9074511239605565, + "mean_feedback_forward_relative_error": 0.42779956998363616, + "min_feedback_forward_cosine": 0.8330364227294922 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036113212761654206, + "instruction_rms": 0.003611321276892142, + "raw_apical_rms": 0.06890865489760267, + "teaching_rms": 0.0036113212761654206, + "traffic_rms": 0.06881233568903938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 6.138535724176108e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5718360645139188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015866967229347967, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3139541429710388 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5707038044929504, + "mean_feedback_forward_cosine": 0.9142736869473611, + "mean_feedback_forward_relative_error": 0.4115029477304028, + "min_feedback_forward_cosine": 0.8446475863456726 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035934747816962773, + "instruction_rms": 0.0035934747830829285, + "raw_apical_rms": 0.06914278411479509, + "teaching_rms": 0.0035934747816962773, + "traffic_rms": 0.06904740347797378 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 6.275681130318844e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5777951293001463e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015709981713933067, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3036772249507904 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5497543215751648, + "mean_feedback_forward_cosine": 0.9204997343401755, + "mean_feedback_forward_relative_error": 0.39606341575422593, + "min_feedback_forward_cosine": 0.8556846976280212 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003602066128294148, + "instruction_rms": 0.0036020661284273583, + "raw_apical_rms": 0.06942569318792734, + "teaching_rms": 0.003602066128294148, + "traffic_rms": 0.0693306285503332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 5.555501658704998e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5659540354013743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01539304192058377, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.2984672192764282 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5293107032775879, + "mean_feedback_forward_cosine": 0.9263052536595252, + "mean_feedback_forward_relative_error": 0.3811599905452421, + "min_feedback_forward_cosine": 0.8660324215888977 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003567323708320373, + "instruction_rms": 0.00356732370952794, + "raw_apical_rms": 0.06916096814635381, + "teaching_rms": 0.003567323708320373, + "traffic_rms": 0.06906711680888128 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 5.934568392262918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5746689225887976e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01555602235159259, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.294402046585083 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5099024772644043, + "mean_feedback_forward_cosine": 0.9318184948736622, + "mean_feedback_forward_relative_error": 0.36650017144218566, + "min_feedback_forward_cosine": 0.8754812479019165 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003629055702909361, + "instruction_rms": 0.003629055701576995, + "raw_apical_rms": 0.06963092681279068, + "teaching_rms": 0.003629055702909361, + "traffic_rms": 0.06953449810498749 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 8.656269301354769e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5694340866787732e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015121122219050057, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28934790353775025 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4909377694129944, + "mean_feedback_forward_cosine": 0.936896110734632, + "mean_feedback_forward_relative_error": 0.35241030973772847, + "min_feedback_forward_cosine": 0.8843867778778076 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003595609623953823, + "instruction_rms": 0.003595609624316693, + "raw_apical_rms": 0.0698145028916054, + "teaching_rms": 0.003595609623953823, + "traffic_rms": 0.06971977788686679 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 6.556560094850283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.565106171226261e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015467650889831446, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28395292448043824 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.47229552268981934, + "mean_feedback_forward_cosine": 0.941606746565911, + "mean_feedback_forward_relative_error": 0.338891624683334, + "min_feedback_forward_cosine": 0.892859935760498 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003570051048288169, + "instruction_rms": 0.0035700510476088967, + "raw_apical_rms": 0.07024312146693426, + "teaching_rms": 0.003570051048288169, + "traffic_rms": 0.07015033299946816 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 5.7252890428571845e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5736121763862837e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015198169248482737, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.276487537021637 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4546135365962982, + "mean_feedback_forward_cosine": 0.9458679460710094, + "mean_feedback_forward_relative_error": 0.3261065620087808, + "min_feedback_forward_cosine": 0.9006040692329407 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003568283303870979, + "instruction_rms": 0.0035682833038841824, + "raw_apical_rms": 0.07017832986819403, + "teaching_rms": 0.003568283303870979, + "traffic_rms": 0.07008566460560568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.809811108927533e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5715400077440312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015248935664702336, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26848984317302704 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4374768137931824, + "mean_feedback_forward_cosine": 0.9498625666864456, + "mean_feedback_forward_relative_error": 0.31367341933711884, + "min_feedback_forward_cosine": 0.9077830910682678 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036007854799207767, + "instruction_rms": 0.003600785480386143, + "raw_apical_rms": 0.07004737717378703, + "teaching_rms": 0.0036007854799207767, + "traffic_rms": 0.06995282721579597 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 6.960980414305595e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.562841669749408e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01491243445533156, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26658033462524416 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4211234748363495, + "mean_feedback_forward_cosine": 0.953569890991334, + "mean_feedback_forward_relative_error": 0.3016835357392988, + "min_feedback_forward_cosine": 0.9144378304481506 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035859103841977, + "instruction_rms": 0.0035859103847275513, + "raw_apical_rms": 0.07025713004176318, + "teaching_rms": 0.0035859103841977, + "traffic_rms": 0.0701636167908582 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.169268201323575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.563881850591078e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015135091509715095, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2559234872341156 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.404587984085083, + "mean_feedback_forward_cosine": 0.9570633845944558, + "mean_feedback_forward_relative_error": 0.2899813310753915, + "min_feedback_forward_cosine": 0.9208983182907104 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003623462345362718, + "instruction_rms": 0.0036234623452563995, + "raw_apical_rms": 0.07028596157188986, + "teaching_rms": 0.003623462345362718, + "traffic_rms": 0.07019001727884343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 6.397198148988537e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5664044388585058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014641788890473018, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25576767461776734 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3892500698566437, + "mean_feedback_forward_cosine": 0.960218318047062, + "mean_feedback_forward_relative_error": 0.2790132438944232, + "min_feedback_forward_cosine": 0.9266842007637024 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003579824925072167, + "instruction_rms": 0.0035798249259937926, + "raw_apical_rms": 0.07039205688694457, + "teaching_rms": 0.003579824925072167, + "traffic_rms": 0.070298913649722 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 8.090054137710467e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5668529465423474e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014838895349049018, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2515813309431076 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37388598918914795, + "mean_feedback_forward_cosine": 0.96321480120382, + "mean_feedback_forward_relative_error": 0.26816824823617935, + "min_feedback_forward_cosine": 0.9322986602783203 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003604920150045678, + "instruction_rms": 0.003604920149207569, + "raw_apical_rms": 0.07069537659520468, + "teaching_rms": 0.003604920150045678, + "traffic_rms": 0.07060154749305185 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.07327636555965e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5603341621615057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014696763425095934, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25099168491363527 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35926011204719543, + "mean_feedback_forward_cosine": 0.9659601334602602, + "mean_feedback_forward_relative_error": 0.2578650359665194, + "min_feedback_forward_cosine": 0.9374220371246338 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036139582229496527, + "instruction_rms": 0.003613958222213897, + "raw_apical_rms": 0.0707432196080217, + "teaching_rms": 0.0036139582229496527, + "traffic_rms": 0.07064876743849158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.36420204816568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.557482387522224e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013753166593612634, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2449511640739441 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3455997407436371, + "mean_feedback_forward_cosine": 0.9684372140515235, + "mean_feedback_forward_relative_error": 0.24818992422473046, + "min_feedback_forward_cosine": 0.9419872760772705 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003618526539846824, + "instruction_rms": 0.0036185265395190084, + "raw_apical_rms": 0.0712440779671978, + "teaching_rms": 0.003618526539846824, + "traffic_rms": 0.07115006688456978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.842152882631126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5608908939363994e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01463015384333008, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24492750885009765 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33338403701782227, + "mean_feedback_forward_cosine": 0.970689308258795, + "mean_feedback_forward_relative_error": 0.2390509692411269, + "min_feedback_forward_cosine": 0.9460171461105347 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003514273819638191, + "instruction_rms": 0.003514273819840301, + "raw_apical_rms": 0.07185023575879976, + "teaching_rms": 0.003514273819638191, + "traffic_rms": 0.07176248139000149 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 5.988686524460718e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5559257207111566e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014423459384459852, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2303639947128296 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32096561789512634, + "mean_feedback_forward_cosine": 0.9729102811505718, + "mean_feedback_forward_relative_error": 0.22972243399389328, + "min_feedback_forward_cosine": 0.9501440525054932 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036082622957538165, + "instruction_rms": 0.0036082622961325665, + "raw_apical_rms": 0.07153499119852696, + "teaching_rms": 0.0036082622957538165, + "traffic_rms": 0.07144157740120281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485987603664398, + "maximum_absolute_post_projection_soma_slope": 5.910189670998989e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5533174637422783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014583604056094006, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.2331293714237213 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31002968549728394, + "mean_feedback_forward_cosine": 0.9748779881385065, + "mean_feedback_forward_relative_error": 0.22114136718934582, + "min_feedback_forward_cosine": 0.9534217119216919 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035347714829293084, + "instruction_rms": 0.0035347714822814118, + "raw_apical_rms": 0.07183233493973475, + "teaching_rms": 0.0035347714829293084, + "traffic_rms": 0.07174284111941574 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 6.25932727871259e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5474592058844107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014028240070744309, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2342703471660614 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29950159788131714, + "mean_feedback_forward_cosine": 0.9767794339887558, + "mean_feedback_forward_relative_error": 0.21252780575906077, + "min_feedback_forward_cosine": 0.9565247893333435 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003610217123786374, + "instruction_rms": 0.0036102171238047123, + "raw_apical_rms": 0.07235855988277147, + "teaching_rms": 0.003610217123786374, + "traffic_rms": 0.07226617896469169 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.098534382521393e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5479094203370572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014232548533931105, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22992673044204712 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28973188996315, + "mean_feedback_forward_cosine": 0.9784697755690543, + "mean_feedback_forward_relative_error": 0.204577540918704, + "min_feedback_forward_cosine": 0.9593740105628967 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035683437195230664, + "instruction_rms": 0.003568343718641344, + "raw_apical_rms": 0.07251229130585289, + "teaching_rms": 0.0035683437195230664, + "traffic_rms": 0.07242229310682763 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 1.0579122289300358e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5491293110287204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013873516771714936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22220664313316346 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2802804708480835, + "mean_feedback_forward_cosine": 0.9800236282810089, + "mean_feedback_forward_relative_error": 0.1969566575942501, + "min_feedback_forward_cosine": 0.9619824886322021 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003545573737626911, + "instruction_rms": 0.0035455737376359816, + "raw_apical_rms": 0.07271454910673958, + "teaching_rms": 0.003545573737626911, + "traffic_rms": 0.07262562983843843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 6.769742810774915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5573255268973508e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013970789926037681, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.2209553227996826 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27100589871406555, + "mean_feedback_forward_cosine": 0.9814942479133606, + "mean_feedback_forward_relative_error": 0.18949198891078273, + "min_feedback_forward_cosine": 0.9644244313240051 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035538451028273837, + "instruction_rms": 0.003553845102324246, + "raw_apical_rms": 0.07255014547313235, + "teaching_rms": 0.0035538451028273837, + "traffic_rms": 0.07246072808047117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.043365795629143e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5467154917628083e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014452166843817482, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2208089984035492 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26140815019607544, + "mean_feedback_forward_cosine": 0.982880175113678, + "mean_feedback_forward_relative_error": 0.18220968784824496, + "min_feedback_forward_cosine": 0.9668254852294922 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035411378733027445, + "instruction_rms": 0.0035411378738394307, + "raw_apical_rms": 0.07286733575368119, + "teaching_rms": 0.0035411378733027445, + "traffic_rms": 0.07277881354938386 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.648495336023188e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5555301059849515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014634393083647996, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21660668154716492 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2515721023082733, + "mean_feedback_forward_cosine": 0.9841807599990599, + "mean_feedback_forward_relative_error": 0.17512519321134012, + "min_feedback_forward_cosine": 0.9692121744155884 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003528841659097951, + "instruction_rms": 0.003528841659134414, + "raw_apical_rms": 0.07273094495250207, + "teaching_rms": 0.003528841659097951, + "traffic_rms": 0.07264293644618175 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.333095197736839e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5532725380837097e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014309781620687424, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2198983741378784 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2422412931919098, + "mean_feedback_forward_cosine": 0.9853565500628564, + "mean_feedback_forward_relative_error": 0.16845456582884635, + "min_feedback_forward_cosine": 0.9714775085449219 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00356773510432647, + "instruction_rms": 0.003567735103613113, + "raw_apical_rms": 0.07288641617085317, + "teaching_rms": 0.00356773510432647, + "traffic_rms": 0.07279675240664363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.341967827618646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5481388218298634e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013792453886977802, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21505139840602874 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2331939935684204, + "mean_feedback_forward_cosine": 0.9864503952764696, + "mean_feedback_forward_relative_error": 0.16202208147414268, + "min_feedback_forward_cosine": 0.9735477566719055 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035382228881439665, + "instruction_rms": 0.0035382228874618767, + "raw_apical_rms": 0.07342302901155884, + "teaching_rms": 0.0035382228881439665, + "traffic_rms": 0.07333512088777719 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.433903531113174e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5485429765373552e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014346749791625278, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2117636529159546 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22525061666965485, + "mean_feedback_forward_cosine": 0.987417588310857, + "mean_feedback_forward_relative_error": 0.15608877027707715, + "min_feedback_forward_cosine": 0.9753058552742004 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003467811643160292, + "instruction_rms": 0.0034678116421256577, + "raw_apical_rms": 0.0732651516126636, + "teaching_rms": 0.003467811643160292, + "traffic_rms": 0.07318059111933263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.541247271092288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543768711456422e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014085245610821483, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2025717242884636 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21740497648715973, + "mean_feedback_forward_cosine": 0.9883876289090803, + "mean_feedback_forward_relative_error": 0.1498993553461567, + "min_feedback_forward_cosine": 0.9769393801689148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036306974563469294, + "instruction_rms": 0.0036306974575807606, + "raw_apical_rms": 0.07310654396514574, + "teaching_rms": 0.0036306974563469294, + "traffic_rms": 0.07301367906981819 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.591091216203495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545394036883777e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015027317682516157, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2066680376958847 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20983608067035675, + "mean_feedback_forward_cosine": 0.9892710851084802, + "mean_feedback_forward_relative_error": 0.1440406883916547, + "min_feedback_forward_cosine": 0.978495180606842 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036078305681864972, + "instruction_rms": 0.003607830568422387, + "raw_apical_rms": 0.07326588468645093, + "teaching_rms": 0.0036078305681864972, + "traffic_rms": 0.07317474353022814 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.03603441568157e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5529996630176128e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015480392477072011, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.20970831125259398 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20208847522735596, + "mean_feedback_forward_cosine": 0.9900660899377638, + "mean_feedback_forward_relative_error": 0.13858351483941078, + "min_feedback_forward_cosine": 0.9800778031349182 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035191287390104386, + "instruction_rms": 0.003519128739096533, + "raw_apical_rms": 0.07320936943679052, + "teaching_rms": 0.0035191287390104386, + "traffic_rms": 0.07312225658221999 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.146664772022632e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5475826626873828e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015389471798350141, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.20343002840995789 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19537170231342316, + "mean_feedback_forward_cosine": 0.9907822551265839, + "mean_feedback_forward_relative_error": 0.13345198249143939, + "min_feedback_forward_cosine": 0.9813547134399414 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034957594861710862, + "instruction_rms": 0.0034957594880671693, + "raw_apical_rms": 0.0738828588731777, + "teaching_rms": 0.0034957594861710862, + "traffic_rms": 0.07379779846383361 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.070416359993942e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542990011263659e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014873055506304843, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19964063883781433 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1883355677127838, + "mean_feedback_forward_cosine": 0.991474203525051, + "mean_feedback_forward_relative_error": 0.12832156977345865, + "min_feedback_forward_cosine": 0.9826860427856445 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035423250416894317, + "instruction_rms": 0.0035423250424304704, + "raw_apical_rms": 0.07414406138734056, + "teaching_rms": 0.0035423250416894317, + "traffic_rms": 0.07405697565431615 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.433654130612922e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5484133176821375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014601856314887793, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19996760390281676 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18150240182876587, + "mean_feedback_forward_cosine": 0.9921058416366577, + "mean_feedback_forward_relative_error": 0.12345341352685806, + "min_feedback_forward_cosine": 0.9839038848876953 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035368324558972893, + "instruction_rms": 0.003536832455649132, + "raw_apical_rms": 0.07443246141932693, + "teaching_rms": 0.0035368324558972893, + "traffic_rms": 0.07434611176159013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.004032839091451e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5400697112334856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014279220159043305, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19564476159572602 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1747015118598938, + "mean_feedback_forward_cosine": 0.9927121670015396, + "mean_feedback_forward_relative_error": 0.1185920762438928, + "min_feedback_forward_cosine": 0.9850529432296753 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035915274110740816, + "instruction_rms": 0.003591527411796789, + "raw_apical_rms": 0.07430636998145877, + "teaching_rms": 0.0035915274110740816, + "traffic_rms": 0.07421717398528853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.405348322388818e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5461576666564704e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013864318215609397, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19875957082271575 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16783694922924042, + "mean_feedback_forward_cosine": 0.9932740196104972, + "mean_feedback_forward_relative_error": 0.11392342960161547, + "min_feedback_forward_cosine": 0.986203134059906 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003583734185415176, + "instruction_rms": 0.0035837341850378382, + "raw_apical_rms": 0.07424991043734266, + "teaching_rms": 0.003583734185415176, + "traffic_rms": 0.07416121185209427 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.0125791050086264e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5419158141631458e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013878449428116037, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.20260698585510253 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1618281751871109, + "mean_feedback_forward_cosine": 0.9937629680479726, + "mean_feedback_forward_relative_error": 0.10969174124540822, + "min_feedback_forward_cosine": 0.9871610999107361 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003478286763943438, + "instruction_rms": 0.0034782867641409117, + "raw_apical_rms": 0.07455198519834542, + "teaching_rms": 0.003478286763943438, + "traffic_rms": 0.07446807860912723 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.071890949333465e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542102533768073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013162751639562506, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18915049114227295 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1558232605457306, + "mean_feedback_forward_cosine": 0.9942391476323528, + "mean_feedback_forward_relative_error": 0.10540983477427114, + "min_feedback_forward_cosine": 0.9880666732788086 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003535855358245179, + "instruction_rms": 0.003535855358514941, + "raw_apical_rms": 0.07485005451974064, + "teaching_rms": 0.003535855358245179, + "traffic_rms": 0.07476403673636119 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.411644276089646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544305088672089e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013818985932131272, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.19254500077724457 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14955919981002808, + "mean_feedback_forward_cosine": 0.9946779704863026, + "mean_feedback_forward_relative_error": 0.1013064254676142, + "min_feedback_forward_cosine": 0.9890223145484924 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003562531528604123, + "instruction_rms": 0.0035625315279557363, + "raw_apical_rms": 0.0746798395252637, + "teaching_rms": 0.003562531528604123, + "traffic_rms": 0.07459210094809259 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0262327521104453e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537492892285799e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014652329516216799, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.19360217828273774 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14410655200481415, + "mean_feedback_forward_cosine": 0.9950764986776537, + "mean_feedback_forward_relative_error": 0.09742516832005593, + "min_feedback_forward_cosine": 0.9898029565811157 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035676243786786333, + "instruction_rms": 0.003567624377833858, + "raw_apical_rms": 0.07492166847924847, + "teaching_rms": 0.0035676243786786333, + "traffic_rms": 0.07483419516109906 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.406086576291273e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537579362513058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014646023011074276, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.19030395037651063 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1386268138885498, + "mean_feedback_forward_cosine": 0.9954412310354172, + "mean_feedback_forward_relative_error": 0.093738068255686, + "min_feedback_forward_cosine": 0.9905557632446289 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035199649503759828, + "instruction_rms": 0.003519964950999063, + "raw_apical_rms": 0.07479722146273009, + "teaching_rms": 0.0035199649503759828, + "traffic_rms": 0.07471181997219575 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.486156926641343e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544761026237976e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01457558466425648, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.1909268178319931 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13337312638759613, + "mean_feedback_forward_cosine": 0.9957873455939754, + "mean_feedback_forward_relative_error": 0.09010233718060678, + "min_feedback_forward_cosine": 0.9912433624267578 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035849803237658047, + "instruction_rms": 0.003584980325426957, + "raw_apical_rms": 0.07548526364493609, + "teaching_rms": 0.0035849803237658047, + "traffic_rms": 0.07539773625772524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.101554354629116e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5268856084482808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014169535282067214, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.19319603216171263 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12824121117591858, + "mean_feedback_forward_cosine": 0.9961044346132586, + "mean_feedback_forward_relative_error": 0.08663714707138077, + "min_feedback_forward_cosine": 0.991897702217102 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035324138877987567, + "instruction_rms": 0.0035324138882821374, + "raw_apical_rms": 0.07569603700592356, + "teaching_rms": 0.0035324138877987567, + "traffic_rms": 0.07561110523697553 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.58067190026668e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5334213973492087e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014581693564766853, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.18684907359600067 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12333238124847412, + "mean_feedback_forward_cosine": 0.9963936132769431, + "mean_feedback_forward_relative_error": 0.08334574962575589, + "min_feedback_forward_cosine": 0.9924991130828857 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003496589974264169, + "instruction_rms": 0.0034965899747340617, + "raw_apical_rms": 0.07597036493797722, + "teaching_rms": 0.003496589974264169, + "traffic_rms": 0.07588712076219775 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.082046321078451e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5395833015813246e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0139032772775167, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.18405272352218627 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11874133348464966, + "mean_feedback_forward_cosine": 0.9966578425899628, + "mean_feedback_forward_relative_error": 0.08022083089716973, + "min_feedback_forward_cosine": 0.9930402636528015 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003513683101232733, + "instruction_rms": 0.0035136831006248627, + "raw_apical_rms": 0.07609684610219249, + "teaching_rms": 0.003513683101232733, + "traffic_rms": 0.07601334137741342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.90832154709642e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5242490335526098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014270058077524241, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.18282177020549775 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11410302668809891, + "mean_feedback_forward_cosine": 0.9969153981054982, + "mean_feedback_forward_relative_error": 0.07707575161851221, + "min_feedback_forward_cosine": 0.9935793280601501 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036083476381545535, + "instruction_rms": 0.0036083476378228592, + "raw_apical_rms": 0.07622372732885632, + "teaching_rms": 0.0036083476381545535, + "traffic_rms": 0.07613545448242943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 6.938371654996445e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5296652058797503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014754368099530188, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.18329295968055725 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10950884222984314, + "mean_feedback_forward_cosine": 0.9971511325528545, + "mean_feedback_forward_relative_error": 0.07406534020218157, + "min_feedback_forward_cosine": 0.9940752387046814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003547374439442843, + "instruction_rms": 0.0035473744401827334, + "raw_apical_rms": 0.07605307224422006, + "teaching_rms": 0.003547374439442843, + "traffic_rms": 0.07596778447124998 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.937065132386124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5315695191526278e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014888307990096316, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.18906341844081878 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10573495179414749, + "mean_feedback_forward_cosine": 0.9973579760520689, + "mean_feedback_forward_relative_error": 0.07130478100190239, + "min_feedback_forward_cosine": 0.9944680333137512 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035124249429167513, + "instruction_rms": 0.0035124249436087693, + "raw_apical_rms": 0.07623972737135004, + "teaching_rms": 0.0035124249429167513, + "traffic_rms": 0.0761559267857974 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.0869854349948582e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5236234503629538e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014453972743486226, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17535344400882721 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1018197163939476, + "mean_feedback_forward_cosine": 0.9975589552233296, + "mean_feedback_forward_relative_error": 0.06853668367670428, + "min_feedback_forward_cosine": 0.9948732852935791 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035406117729962503, + "instruction_rms": 0.0035406117724872217, + "raw_apical_rms": 0.07615759788290066, + "teaching_rms": 0.0035406117729962503, + "traffic_rms": 0.07607255346579143 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 9.821511071095301e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5451872344122994e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015035056867348424, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1841877987718582 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09843720495700836, + "mean_feedback_forward_cosine": 0.9977367097331632, + "mean_feedback_forward_relative_error": 0.06598064537730909, + "min_feedback_forward_cosine": 0.9952043294906616 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003495830509761125, + "instruction_rms": 0.0034958305109188876, + "raw_apical_rms": 0.07629823092094715, + "teaching_rms": 0.003495830509761125, + "traffic_rms": 0.07621522743955157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1210785544335522e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5239736843174995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015060494768254747, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.17492715606689452 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09479423612356186, + "mean_feedback_forward_cosine": 0.9979092690252489, + "mean_feedback_forward_relative_error": 0.0634160170992536, + "min_feedback_forward_cosine": 0.9955540895462036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003531479361941745, + "instruction_rms": 0.0035314793627695493, + "raw_apical_rms": 0.0763427653959354, + "teaching_rms": 0.003531479361941745, + "traffic_rms": 0.07625820705492264 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.1792678122901634e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.519641786309308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01462348411231385, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1827969622707367 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09126103669404984, + "mean_feedback_forward_cosine": 0.9980602264404297, + "mean_feedback_forward_relative_error": 0.06107467719383778, + "min_feedback_forward_cosine": 0.9958809614181519 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003511989402995332, + "instruction_rms": 0.0035119894045005503, + "raw_apical_rms": 0.07645319287249783, + "teaching_rms": 0.003511989402995332, + "traffic_rms": 0.07636960366053744 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.13380873157621e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.522335222133876e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015485400517680681, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.17390078784942628 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08786570280790329, + "mean_feedback_forward_cosine": 0.9982036775158297, + "mean_feedback_forward_relative_error": 0.058767640722855445, + "min_feedback_forward_cosine": 0.9961830377578735 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003510973985424096, + "instruction_rms": 0.0035109739863188965, + "raw_apical_rms": 0.07643620183467235, + "teaching_rms": 0.003510973985424096, + "traffic_rms": 0.07635283791254978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.41351726674111e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529499990270176e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014838685674348297, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.17431724754810332 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08470561355352402, + "mean_feedback_forward_cosine": 0.9983381205989469, + "mean_feedback_forward_relative_error": 0.05651552302222098, + "min_feedback_forward_cosine": 0.9964495897293091 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035387833308349556, + "instruction_rms": 0.0035387833295479386, + "raw_apical_rms": 0.07682818089801362, + "teaching_rms": 0.0035387833308349556, + "traffic_rms": 0.07674391562968257 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 7.734478657539512e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.525725334664447e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014106072571530726, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1742467817211151 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08137988299131393, + "mean_feedback_forward_cosine": 0.9984661398395416, + "mean_feedback_forward_relative_error": 0.05429629410707181, + "min_feedback_forward_cosine": 0.996722936630249 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035368698563421896, + "instruction_rms": 0.0035368698565948807, + "raw_apical_rms": 0.07668344871325335, + "teaching_rms": 0.0035368698563421896, + "traffic_rms": 0.07659905406782085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.479236868197404e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5298777851516374e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014231159298908364, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.17773329139709473 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07825799286365509, + "mean_feedback_forward_cosine": 0.998579828969894, + "mean_feedback_forward_relative_error": 0.05224089318465802, + "min_feedback_forward_cosine": 0.9969693422317505 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034274400758817418, + "instruction_rms": 0.0034274400757067897, + "raw_apical_rms": 0.07701370861979899, + "teaching_rms": 0.0034274400758817418, + "traffic_rms": 0.07693449183067701 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 1.1169006342015564e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5298715934851535e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015109962188135194, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1731295151090622 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0752711072564125, + "mean_feedback_forward_cosine": 0.9986858887057151, + "mean_feedback_forward_relative_error": 0.05024578530461558, + "min_feedback_forward_cosine": 0.9971962571144104 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034798355626603264, + "instruction_rms": 0.0034798355619360087, + "raw_apical_rms": 0.0775893310921717, + "teaching_rms": 0.0034798355626603264, + "traffic_rms": 0.07750835716591364 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0516922088754654e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5225252800836012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015427078382848262, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.17347687585830687 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.072512187063694, + "mean_feedback_forward_cosine": 0.9987837056959828, + "mean_feedback_forward_relative_error": 0.048337063541816126, + "min_feedback_forward_cosine": 0.9973939061164856 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003445245711270811, + "instruction_rms": 0.003445245710523916, + "raw_apical_rms": 0.07776831356730385, + "teaching_rms": 0.003445245711270811, + "traffic_rms": 0.0776889849298262 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.540880796521378e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5202650409199263e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015316771343777411, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.17246941825151443 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06970735639333725, + "mean_feedback_forward_cosine": 0.9988761851864476, + "mean_feedback_forward_relative_error": 0.04646128998889077, + "min_feedback_forward_cosine": 0.9975912570953369 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034676542428025547, + "instruction_rms": 0.003467654242923529, + "raw_apical_rms": 0.0780375994479244, + "teaching_rms": 0.0034676542428025547, + "traffic_rms": 0.07795783358289698 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.036554522215738e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5187794423048757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014605065470960913, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1780101472902298 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06705942004919052, + "mean_feedback_forward_cosine": 0.9989592279157331, + "mean_feedback_forward_relative_error": 0.04471062772697018, + "min_feedback_forward_cosine": 0.9977735280990601 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034269871487030285, + "instruction_rms": 0.0034269871481415267, + "raw_apical_rms": 0.07826105492704556, + "teaching_rms": 0.0034269871487030285, + "traffic_rms": 0.0781829204866453 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.952651515097386e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5215194881173365e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01564781421400523, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.17068311315059662 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06453179568052292, + "mean_feedback_forward_cosine": 0.9990373100003889, + "mean_feedback_forward_relative_error": 0.04299624738914351, + "min_feedback_forward_cosine": 0.997937798500061 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003464505163004312, + "instruction_rms": 0.003464505162830319, + "raw_apical_rms": 0.07820298580457843, + "teaching_rms": 0.003464505163004312, + "traffic_rms": 0.07812343209678141 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.397366002554918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5236947503659647e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015184065074857188, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.16956559858322143 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.062339261174201965, + "mean_feedback_forward_cosine": 0.999108495250825, + "mean_feedback_forward_relative_error": 0.04137106588290584, + "min_feedback_forward_cosine": 0.998076319694519 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034532588682173603, + "instruction_rms": 0.00345325886788395, + "raw_apical_rms": 0.07829725195512172, + "teaching_rms": 0.0034532588682173603, + "traffic_rms": 0.07821822696815829 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.947472534577173e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5161306306876885e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01527853877513493, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.17076773929595948 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059681862592697144, + "mean_feedback_forward_cosine": 0.9991771847971024, + "mean_feedback_forward_relative_error": 0.039746583319239075, + "min_feedback_forward_cosine": 0.9982354640960693 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035120564590859003, + "instruction_rms": 0.003512056458114103, + "raw_apical_rms": 0.07870747112464983, + "teaching_rms": 0.0035120564590859003, + "traffic_rms": 0.07862635169707961 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.217246343951047e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5232262780781532e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015276575791907918, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.17142115005493164 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05723540857434273, + "mean_feedback_forward_cosine": 0.9992400484700357, + "mean_feedback_forward_relative_error": 0.038202963408923915, + "min_feedback_forward_cosine": 0.9983772039413452 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003494858156359144, + "instruction_rms": 0.003494858156024841, + "raw_apical_rms": 0.07849947800043329, + "teaching_rms": 0.003494858156359144, + "traffic_rms": 0.07841873976114817 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1524583243271991e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.519435529878183e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01421604195163046, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.17196555834770202 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05511709302663803, + "mean_feedback_forward_cosine": 0.9992951423891129, + "mean_feedback_forward_relative_error": 0.036785399961856105, + "min_feedback_forward_cosine": 0.9984930157661438 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0033877543881544784, + "instruction_rms": 0.0033877543891387193, + "raw_apical_rms": 0.07872908532779466, + "teaching_rms": 0.0033877543881544784, + "traffic_rms": 0.07865335545989838 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.518948473896671e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5275384610118153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014686607838940357, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.16454553032398223 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.053032949566841125, + "mean_feedback_forward_cosine": 0.9993482866594868, + "mean_feedback_forward_relative_error": 0.035372225267271844, + "min_feedback_forward_cosine": 0.9986045360565186 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034628492072512037, + "instruction_rms": 0.0034628492076276113, + "raw_apical_rms": 0.07910306808613789, + "teaching_rms": 0.0034628492072512037, + "traffic_rms": 0.07902444817274976 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.613498891918425e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.524672210198651e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014690587823124638, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.1659625129032135 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0507659912109375, + "mean_feedback_forward_cosine": 0.9993979853968467, + "mean_feedback_forward_relative_error": 0.03399836599466301, + "min_feedback_forward_cosine": 0.9987220764160156 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003486906064305537, + "instruction_rms": 0.0034869060650725967, + "raw_apical_rms": 0.0789895851124582, + "teaching_rms": 0.003486906064305537, + "traffic_rms": 0.07890988116910472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.605545076534327e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5197526043836565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014318585798445338, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.16840963133335113 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04880470037460327, + "mean_feedback_forward_cosine": 0.9994427580987254, + "mean_feedback_forward_relative_error": 0.032710261913316864, + "min_feedback_forward_cosine": 0.9988191723823547 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034286222954502533, + "instruction_rms": 0.0034286222948842304, + "raw_apical_rms": 0.0791167926077213, + "teaching_rms": 0.0034286222954502533, + "traffic_rms": 0.07903956783426677 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.1512290853943341e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5215864465985768e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015061459454309834, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.16472424047470094 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.046967122703790665, + "mean_feedback_forward_cosine": 0.9994842486996804, + "mean_feedback_forward_relative_error": 0.03146700993660958, + "min_feedback_forward_cosine": 0.9989062547683716 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034543203241874623, + "instruction_rms": 0.003454320323692447, + "raw_apical_rms": 0.07911317919499969, + "teaching_rms": 0.0034543203241874623, + "traffic_rms": 0.07903508655923748 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.277758294956584e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513490919059114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015046410993624913, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1640119525885582 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04525262862443924, + "mean_feedback_forward_cosine": 0.9995235320060484, + "mean_feedback_forward_relative_error": 0.030242789476629225, + "min_feedback_forward_cosine": 0.9989849328994751 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003505403215488739, + "instruction_rms": 0.00350540321593002, + "raw_apical_rms": 0.07912524233391424, + "teaching_rms": 0.003505403215488739, + "traffic_rms": 0.07904428170793545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.812236356765425e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5169537672197523e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015159762437062213, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.166050426030159 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04345744103193283, + "mean_feedback_forward_cosine": 0.9995605657177586, + "mean_feedback_forward_relative_error": 0.02904382453209931, + "min_feedback_forward_cosine": 0.9990651607513428 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035393196707221674, + "instruction_rms": 0.0035393196696472083, + "raw_apical_rms": 0.07895586536078089, + "teaching_rms": 0.0035393196707221674, + "traffic_rms": 0.0788733693364956 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.617264768417954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5167174494917777e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015240176688414445, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.1662076429438591 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04176519811153412, + "mean_feedback_forward_cosine": 0.9995943346331196, + "mean_feedback_forward_relative_error": 0.027905780191142714, + "min_feedback_forward_cosine": 0.9991357326507568 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003516851113984911, + "instruction_rms": 0.003516851114126374, + "raw_apical_rms": 0.07934237251546009, + "teaching_rms": 0.003516851113984911, + "traffic_rms": 0.07926160164845057 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.927081213665588e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5272991444347993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015187137801153352, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.17011467378497125 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04019809141755104, + "mean_feedback_forward_cosine": 0.9996240754281321, + "mean_feedback_forward_relative_error": 0.026859281975175103, + "min_feedback_forward_cosine": 0.9991989135742188 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003448810612157141, + "instruction_rms": 0.0034488106128691583, + "raw_apical_rms": 0.0796090302692927, + "teaching_rms": 0.003448810612157141, + "traffic_rms": 0.07953194447727764 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.7103857399361e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.523996955379413e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01492849317895086, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.16408861306190492 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038644999265670776, + "mean_feedback_forward_cosine": 0.9996531048128682, + "mean_feedback_forward_relative_error": 0.025800723402250196, + "min_feedback_forward_cosine": 0.9992592930793762 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035202597556042037, + "instruction_rms": 0.0035202597556102436, + "raw_apical_rms": 0.07971225660496446, + "teaching_rms": 0.0035202597556042037, + "traffic_rms": 0.07963165417954865 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.1783674125354082e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5197906453247196e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01535256807704044, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.16387008115291596 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03719918057322502, + "mean_feedback_forward_cosine": 0.9996788405602978, + "mean_feedback_forward_relative_error": 0.024823390878736973, + "min_feedback_forward_cosine": 0.9993136525154114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034862356175156812, + "instruction_rms": 0.0034862356161719844, + "raw_apical_rms": 0.0796233196621906, + "teaching_rms": 0.0034862356175156812, + "traffic_rms": 0.07954412217449612 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0124649207909897e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5100573789575725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015956955036797622, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.16269615334510804 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.035789620131254196, + "mean_feedback_forward_cosine": 0.9997035611060358, + "mean_feedback_forward_relative_error": 0.023847881163800915, + "min_feedback_forward_cosine": 0.9993649125099182 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035357990362742857, + "instruction_rms": 0.003535799037342637, + "raw_apical_rms": 0.07968405260724283, + "teaching_rms": 0.0035357990362742857, + "traffic_rms": 0.07960272501545694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0262154148676927e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512815379923576e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014595247023227986, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.16219510801792145 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03450062870979309, + "mean_feedback_forward_cosine": 0.999725676351978, + "mean_feedback_forward_relative_error": 0.022937791602265452, + "min_feedback_forward_cosine": 0.9994091987609863 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034607600347872436, + "instruction_rms": 0.0034607600338873563, + "raw_apical_rms": 0.08000673016027257, + "teaching_rms": 0.0034607600347872436, + "traffic_rms": 0.07992918212158623 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.115185406471937e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5132679730621306e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015766540240688, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.16312291798114775 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03321545943617821, + "mean_feedback_forward_cosine": 0.9997461188224054, + "mean_feedback_forward_relative_error": 0.022066733890002774, + "min_feedback_forward_cosine": 0.9994525909423828 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034335899241077694, + "instruction_rms": 0.0034335899243701077, + "raw_apical_rms": 0.07965220592247597, + "teaching_rms": 0.0034335899241077694, + "traffic_rms": 0.07957544158849365 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.0353105039939692e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5219333090705045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014654316319455121, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.15964929082393647 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03320581838488579, + "mean_feedback_forward_cosine": 0.9997462803317655, + "mean_feedback_forward_relative_error": 0.02205745293007743, + "min_feedback_forward_cosine": 0.9994529485702515 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002524377649175733, + "instruction_rms": 0.002524377648267595, + "raw_apical_rms": 0.07956611442997562, + "teaching_rms": 0.002524377649175733, + "traffic_rms": 0.07952373511436356 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.715528648257532e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509039661375874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014598867267780324, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.0916484384560585 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0332002267241478, + "mean_feedback_forward_cosine": 0.9997463918501331, + "mean_feedback_forward_relative_error": 0.022051295460832697, + "min_feedback_forward_cosine": 0.9994531273841858 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0022497534103624855, + "instruction_rms": 0.002249753409728509, + "raw_apical_rms": 0.07960016190938568, + "teaching_rms": 0.0022497534103624855, + "traffic_rms": 0.07956597747766647 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1159445278963176e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507462771047379e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014711402169480659, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.06731309509515762 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033195290714502335, + "mean_feedback_forward_cosine": 0.9997464976003093, + "mean_feedback_forward_relative_error": 0.022045515192252015, + "min_feedback_forward_cosine": 0.9994533061981201 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002138066689938949, + "instruction_rms": 0.0021380666902328957, + "raw_apical_rms": 0.07967308021131113, + "teaching_rms": 0.002138066689938949, + "traffic_rms": 0.07964159952796075 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.0855472254434062e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.51108435162377e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01468375803516038, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.056721894896030425 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033190567046403885, + "mean_feedback_forward_cosine": 0.9997466033504855, + "mean_feedback_forward_relative_error": 0.02204016044795994, + "min_feedback_forward_cosine": 0.9994533061981201 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002051799172275007, + "instruction_rms": 0.0020517991726182517, + "raw_apical_rms": 0.0796808384156964, + "teaching_rms": 0.002051799172275007, + "traffic_rms": 0.07965182233170502 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.896699282307964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5092704317774955e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01471528238026482, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.04962461613893509 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033185623586177826, + "mean_feedback_forward_cosine": 0.9997466552642084, + "mean_feedback_forward_relative_error": 0.022034939318414656, + "min_feedback_forward_cosine": 0.9994535446166992 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020726900735696325, + "instruction_rms": 0.0020726900734601827, + "raw_apical_rms": 0.07960521261310456, + "teaching_rms": 0.0020726900735696325, + "traffic_rms": 0.07957533626496159 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.0391228499884164e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5078323246202397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014989065416981587, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.04779350155115127 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0331808440387249, + "mean_feedback_forward_cosine": 0.9997467398643494, + "mean_feedback_forward_relative_error": 0.022029850722080278, + "min_feedback_forward_cosine": 0.999453604221344 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002036422250482297, + "instruction_rms": 0.0020364222508524317, + "raw_apical_rms": 0.07962794960976033, + "teaching_rms": 0.002036422250482297, + "traffic_rms": 0.07959921322394772 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.102458875531738e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5105934403930013e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014656352690534264, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.04343980648636818 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033176030963659286, + "mean_feedback_forward_cosine": 0.9997468321554123, + "mean_feedback_forward_relative_error": 0.022024953001809698, + "min_feedback_forward_cosine": 0.9994538426399231 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019692105624990824, + "instruction_rms": 0.001969210562701008, + "raw_apical_rms": 0.07960509195788704, + "teaching_rms": 0.0019692105624990824, + "traffic_rms": 0.07957783316156186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.2611951660801424e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5110189482235556e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0147731701658162, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.0396160158097744 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033171720802783966, + "mean_feedback_forward_cosine": 0.9997469071419008, + "mean_feedback_forward_relative_error": 0.022020161482355288, + "min_feedback_forward_cosine": 0.9994539022445679 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019188334077852975, + "instruction_rms": 0.0019188334085098486, + "raw_apical_rms": 0.07958338833757954, + "teaching_rms": 0.0019188334077852975, + "traffic_rms": 0.07955734412186102 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.26731820779969e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5061211124889174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015009183275247655, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.03591494343280792 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03316723182797432, + "mean_feedback_forward_cosine": 0.9997470052011551, + "mean_feedback_forward_relative_error": 0.02201542094530117, + "min_feedback_forward_cosine": 0.9994542002677917 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019498957651283382, + "instruction_rms": 0.0019498957650978819, + "raw_apical_rms": 0.07959046767113151, + "teaching_rms": 0.0019498957651283382, + "traffic_rms": 0.07956336137100854 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0529499405720344e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.51076287096211e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014459088446631522, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.034892128787636756 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03316301852464676, + "mean_feedback_forward_cosine": 0.9997470724967218, + "mean_feedback_forward_relative_error": 0.022010906599462032, + "min_feedback_forward_cosine": 0.9994543194770813 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018224904877953568, + "instruction_rms": 0.0018224904876616578, + "raw_apical_rms": 0.07956183794329776, + "teaching_rms": 0.0018224904877953568, + "traffic_rms": 0.07953792842441589 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.690877078583071e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.508789933180616e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014563443524614637, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.03149836225405336 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03315827250480652, + "mean_feedback_forward_cosine": 0.9997471436377494, + "mean_feedback_forward_relative_error": 0.022006309305828427, + "min_feedback_forward_cosine": 0.9994544982910156 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019209985393784101, + "instruction_rms": 0.0019209985401233442, + "raw_apical_rms": 0.07959803661440297, + "teaching_rms": 0.0019209985393784101, + "traffic_rms": 0.07957096637679241 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1551588130487289e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5072826531844513e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014283908329078016, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.03162747761368752 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03315358608961105, + "mean_feedback_forward_cosine": 0.9997472359288123, + "mean_feedback_forward_relative_error": 0.02200179961660216, + "min_feedback_forward_cosine": 0.9994546175003052 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018567101702496465, + "instruction_rms": 0.0018567101701632818, + "raw_apical_rms": 0.07956063304666747, + "teaching_rms": 0.0018567101702496465, + "traffic_rms": 0.07953554369900595 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.462548433702977e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.504557128584169e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014699166657594836, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.029898817348778247 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03314914554357529, + "mean_feedback_forward_cosine": 0.9997473224516837, + "mean_feedback_forward_relative_error": 0.02199734953202067, + "min_feedback_forward_cosine": 0.9994547963142395 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018413790513199125, + "instruction_rms": 0.0018413790507320182, + "raw_apical_rms": 0.07955049423748724, + "teaching_rms": 0.0018413790513199125, + "traffic_rms": 0.0795254641147627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.593823253022492e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5111988895415303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014515494039985179, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.028182127690315247 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03314487263560295, + "mean_feedback_forward_cosine": 0.9997473859017895, + "mean_feedback_forward_relative_error": 0.021993035916239023, + "min_feedback_forward_cosine": 0.999454915523529 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017795296092024486, + "instruction_rms": 0.0017795296086548704, + "raw_apical_rms": 0.07955582123128997, + "teaching_rms": 0.0017795296092024486, + "traffic_rms": 0.0795319232146296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0053417298649947e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5073376856272273e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014497252674330037, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.02656099889576435 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03314022719860077, + "mean_feedback_forward_cosine": 0.9997474820383133, + "mean_feedback_forward_relative_error": 0.021988571696584264, + "min_feedback_forward_cosine": 0.9994550347328186 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001797634265611029, + "instruction_rms": 0.0017976342654406304, + "raw_apical_rms": 0.07951707143616796, + "teaching_rms": 0.001797634265611029, + "traffic_rms": 0.07949222952459628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.597399497351944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509047463529293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014660497326169326, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.026259113113284112 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03313567489385605, + "mean_feedback_forward_cosine": 0.9997475435656886, + "mean_feedback_forward_relative_error": 0.02198437445106045, + "min_feedback_forward_cosine": 0.9994550943374634 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017471089053553352, + "instruction_rms": 0.0017471089048102062, + "raw_apical_rms": 0.07946765434164102, + "teaching_rms": 0.0017471089053553352, + "traffic_rms": 0.07944430355378941 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.617651386406578e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5151259222042737e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01445668886878179, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.024145349757671356 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03313136473298073, + "mean_feedback_forward_cosine": 0.9997476185521772, + "mean_feedback_forward_relative_error": 0.021980281438558333, + "min_feedback_forward_cosine": 0.9994553327560425 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016865049301085428, + "instruction_rms": 0.0016865049298738439, + "raw_apical_rms": 0.07948849854382041, + "teaching_rms": 0.0016865049301085428, + "traffic_rms": 0.07946660361099382 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.524128247699991e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5104894937384047e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014659490321058812, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.022238348321318626 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03312728554010391, + "mean_feedback_forward_cosine": 0.9997476896932048, + "mean_feedback_forward_relative_error": 0.021976222794863484, + "min_feedback_forward_cosine": 0.9994553923606873 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016817080548818311, + "instruction_rms": 0.0016817080542188324, + "raw_apical_rms": 0.07951743508161727, + "teaching_rms": 0.0016817080548818311, + "traffic_rms": 0.07949618469505985 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.4558110674825e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.50937748467848e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014636433726206802, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.021405158285796644 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03312275558710098, + "mean_feedback_forward_cosine": 0.99974775122058, + "mean_feedback_forward_relative_error": 0.021972021899156032, + "min_feedback_forward_cosine": 0.9994556307792664 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017309468360352236, + "instruction_rms": 0.0017309468359035642, + "raw_apical_rms": 0.07953350997944694, + "teaching_rms": 0.0017309468360352236, + "traffic_rms": 0.07950999837100513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.687001695761865e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509258159906198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01421481767207631, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.02178585317194462 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03311806544661522, + "mean_feedback_forward_cosine": 0.9997478473571039, + "mean_feedback_forward_relative_error": 0.021967657210846103, + "min_feedback_forward_cosine": 0.9994557499885559 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017604146655849492, + "instruction_rms": 0.001760414665925824, + "raw_apical_rms": 0.07950234198003521, + "teaching_rms": 0.0017604146655849492, + "traffic_rms": 0.07947830958813384 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1629545326741209e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.508939020407699e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015019918875239891, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.021678331478238107 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03311392292380333, + "mean_feedback_forward_cosine": 0.9997479281117839, + "mean_feedback_forward_relative_error": 0.021963649519508886, + "min_feedback_forward_cosine": 0.9994558095932007 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016276494832314413, + "instruction_rms": 0.0016276494832636883, + "raw_apical_rms": 0.07949333313050187, + "teaching_rms": 0.0016276494832314413, + "traffic_rms": 0.07947271086670903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.639353410444528e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5086631254468076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014495193289308194, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.0191543836581707 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033109523355960846, + "mean_feedback_forward_cosine": 0.9997480069437334, + "mean_feedback_forward_relative_error": 0.021959528101668242, + "min_feedback_forward_cosine": 0.9994560480117798 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016840093252409765, + "instruction_rms": 0.0016840093253523928, + "raw_apical_rms": 0.07949006482803656, + "teaching_rms": 0.0016840093252409765, + "traffic_rms": 0.0794677123452571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.4564647667994e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512491338890048e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014758534750771176, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.019814275299310683 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03310495615005493, + "mean_feedback_forward_cosine": 0.9997480992347964, + "mean_feedback_forward_relative_error": 0.021955404190286514, + "min_feedback_forward_cosine": 0.9994560480117798 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016450483733881712, + "instruction_rms": 0.0016450483738384846, + "raw_apical_rms": 0.07945055397437419, + "teaching_rms": 0.0016450483733881712, + "traffic_rms": 0.07942905777135492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.15649328336076e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5097370078311765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014317199746892446, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.0184604742115736 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03310025855898857, + "mean_feedback_forward_cosine": 0.999748170375824, + "mean_feedback_forward_relative_error": 0.021951179891344038, + "min_feedback_forward_cosine": 0.9994562864303589 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016646489682154704, + "instruction_rms": 0.0016646489675116367, + "raw_apical_rms": 0.07939332941780236, + "teaching_rms": 0.0016646489682154704, + "traffic_rms": 0.0793713704016793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1870082516907132e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5113734466973797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014584545773621074, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.018684975422620775 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03309587761759758, + "mean_feedback_forward_cosine": 0.9997482434395821, + "mean_feedback_forward_relative_error": 0.021947216663149097, + "min_feedback_forward_cosine": 0.999456524848938 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015961695537588894, + "instruction_rms": 0.0015961695537925569, + "raw_apical_rms": 0.07936028725061918, + "teaching_rms": 0.0015961695537588894, + "traffic_rms": 0.07933939059594357 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.1308503644613666e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513410244674237e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014335429213391412, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.017586506919264795 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03309120610356331, + "mean_feedback_forward_cosine": 0.9997483414988364, + "mean_feedback_forward_relative_error": 0.021942997816950083, + "min_feedback_forward_cosine": 0.9994565844535828 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016478505165141738, + "instruction_rms": 0.0016478505164579625, + "raw_apical_rms": 0.07938729879437009, + "teaching_rms": 0.0016478505165141738, + "traffic_rms": 0.07936557637829424 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.983491366938324e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5100228695085506e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014283761860934935, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.017436602291464806 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033086344599723816, + "mean_feedback_forward_cosine": 0.9997483972580202, + "mean_feedback_forward_relative_error": 0.021938673716278806, + "min_feedback_forward_cosine": 0.9994568228721619 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001652176430966241, + "instruction_rms": 0.0016521764307322567, + "raw_apical_rms": 0.07941043600527799, + "teaching_rms": 0.001652176430966241, + "traffic_rms": 0.07938796849465479 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0065264888226011e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5165874757897445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014618440249554736, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.017328789977729322 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033081695437431335, + "mean_feedback_forward_cosine": 0.9997484876263526, + "mean_feedback_forward_relative_error": 0.021934578811088876, + "min_feedback_forward_cosine": 0.9994568824768066 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001546432797776313, + "instruction_rms": 0.0015464327977135898, + "raw_apical_rms": 0.07935611624620695, + "teaching_rms": 0.001546432797776313, + "traffic_rms": 0.0793365383409044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.005647345044054e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.512237536886877e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014617593939318123, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.01631776251614094 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03307739645242691, + "mean_feedback_forward_cosine": 0.9997485837628765, + "mean_feedback_forward_relative_error": 0.02193071631594531, + "min_feedback_forward_cosine": 0.999457061290741 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015037357866316053, + "instruction_rms": 0.001503735786111609, + "raw_apical_rms": 0.07932926771432572, + "teaching_rms": 0.0015037357866316053, + "traffic_rms": 0.0793099664185165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0920367543576504e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.511391510659731e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014229565556877086, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.014984703678786754 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03307286649942398, + "mean_feedback_forward_cosine": 0.9997486299084078, + "mean_feedback_forward_relative_error": 0.021926819692335783, + "min_feedback_forward_cosine": 0.9994572997093201 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014736606712644036, + "instruction_rms": 0.0014736606712586367, + "raw_apical_rms": 0.07929485206923978, + "teaching_rms": 0.0014736606712644036, + "traffic_rms": 0.07927646701433155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485996544361115, + "maximum_absolute_post_projection_soma_slope": 1.3293023926053138e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513993364567757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014414970683719124, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.01485530969157815 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033068444579839706, + "mean_feedback_forward_cosine": 0.999748706817627, + "mean_feedback_forward_relative_error": 0.02192302085759659, + "min_feedback_forward_cosine": 0.9994574189186096 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014623085161146114, + "instruction_rms": 0.0014623085164752676, + "raw_apical_rms": 0.07926587091041316, + "teaching_rms": 0.0014623085161146114, + "traffic_rms": 0.07924741455353097 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.738642520711437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5127331979987113e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014329939355551581, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.013986171801686287 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0330636203289032, + "mean_feedback_forward_cosine": 0.9997487856495765, + "mean_feedback_forward_relative_error": 0.02191885963322655, + "min_feedback_forward_cosine": 0.999457597732544 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001488147267169501, + "instruction_rms": 0.0014881472668486888, + "raw_apical_rms": 0.07931577823971268, + "teaching_rms": 0.001488147267169501, + "traffic_rms": 0.07929633046706953 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.666328454864924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510809225351413e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014807647480394788, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.014039978965818881 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03305862098932266, + "mean_feedback_forward_cosine": 0.9997488664042565, + "mean_feedback_forward_relative_error": 0.02191468203560479, + "min_feedback_forward_cosine": 0.9994577765464783 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015287955608723936, + "instruction_rms": 0.0015287955609434537, + "raw_apical_rms": 0.07930816031100323, + "teaching_rms": 0.0015287955608723936, + "traffic_rms": 0.07928856228118013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.03557548781464e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5161297888060356e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014344471194332043, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.014608249745666982 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03305349126458168, + "mean_feedback_forward_cosine": 0.9997489759998937, + "mean_feedback_forward_relative_error": 0.02191058897803868, + "min_feedback_forward_cosine": 0.999457836151123 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00150112125665082, + "instruction_rms": 0.0015011212561185463, + "raw_apical_rms": 0.07934008124301642, + "teaching_rms": 0.00150112125665082, + "traffic_rms": 0.07932065923465888 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.94770408876866e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5148499394904662e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014618811664349857, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.013778267548978328 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033047933131456375, + "mean_feedback_forward_cosine": 0.9997490509863822, + "mean_feedback_forward_relative_error": 0.021906138112347935, + "min_feedback_forward_cosine": 0.9994581937789917 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015349943191689491, + "instruction_rms": 0.001534994320111577, + "raw_apical_rms": 0.07928782792113892, + "teaching_rms": 0.0015349943191689491, + "traffic_rms": 0.07926703245957184 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485989093780518, + "maximum_absolute_post_projection_soma_slope": 1.0586501986153962e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5146327100451497e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014320241728654594, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.014246001432240009 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03304282948374748, + "mean_feedback_forward_cosine": 0.9997491278956013, + "mean_feedback_forward_relative_error": 0.021901976181975296, + "min_feedback_forward_cosine": 0.9994581937789917 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014740610487929146, + "instruction_rms": 0.0014740610494236002, + "raw_apical_rms": 0.0792762921437456, + "teaching_rms": 0.0014740610487929146, + "traffic_rms": 0.07925717504643073 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.784923094002806e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514359236451982e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01417467304159868, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.012947517290711403 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03303813189268112, + "mean_feedback_forward_cosine": 0.9997492124957423, + "mean_feedback_forward_relative_error": 0.021897993125622312, + "min_feedback_forward_cosine": 0.9994583129882812 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014300269964651832, + "instruction_rms": 0.0014300269963151006, + "raw_apical_rms": 0.07922977658315739, + "teaching_rms": 0.0014300269964651832, + "traffic_rms": 0.079211840163229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.412162344484386e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5176864371716052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014720802594932686, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.01280154710650444 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033033035695552826, + "mean_feedback_forward_cosine": 0.9997492855595004, + "mean_feedback_forward_relative_error": 0.021893944891710437, + "min_feedback_forward_cosine": 0.9994584918022156 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001422550292052476, + "instruction_rms": 0.0014225502922694034, + "raw_apical_rms": 0.0792274778439576, + "teaching_rms": 0.001422550292052476, + "traffic_rms": 0.07920960937542025 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1356434015397099e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5133326718182288e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014270647888960066, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.012516646864414215 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033027809113264084, + "mean_feedback_forward_cosine": 0.9997493951551376, + "mean_feedback_forward_relative_error": 0.021889692908453363, + "min_feedback_forward_cosine": 0.9994587898254395 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014556578825081108, + "instruction_rms": 0.001455657882529878, + "raw_apical_rms": 0.0792284660023234, + "teaching_rms": 0.0014556578825081108, + "traffic_rms": 0.07920974768793237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.947395795961711e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5128178433672987e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01479745777508257, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.012854426398873329 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033022891730070114, + "mean_feedback_forward_cosine": 0.9997494701416262, + "mean_feedback_forward_relative_error": 0.02188554198871697, + "min_feedback_forward_cosine": 0.9994589686393738 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014239408222492184, + "instruction_rms": 0.0014239408218267473, + "raw_apical_rms": 0.07920914969742018, + "teaching_rms": 0.0014239408222492184, + "traffic_rms": 0.0791905802278244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.806539935652836e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5097579659283822e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014466917315100964, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.012006430386602878 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03301799297332764, + "mean_feedback_forward_cosine": 0.9997495528190367, + "mean_feedback_forward_relative_error": 0.021881497404988733, + "min_feedback_forward_cosine": 0.9994590282440186 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013793257124983078, + "instruction_rms": 0.0013793257123449474, + "raw_apical_rms": 0.07921858917427291, + "teaching_rms": 0.0013793257124983078, + "traffic_rms": 0.07920085743444595 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.272689993622407e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5135981381251612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014045345101084775, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.011773471644818782 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03301255777478218, + "mean_feedback_forward_cosine": 0.99974964703283, + "mean_feedback_forward_relative_error": 0.021877035844109712, + "min_feedback_forward_cosine": 0.9994592666625977 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014504367673261113, + "instruction_rms": 0.0014504367676894732, + "raw_apical_rms": 0.07922338906626804, + "teaching_rms": 0.0014504367673261113, + "traffic_rms": 0.07920431146452989 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.921401217776292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514224387228165e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014464510775200487, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.012035533401668071 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.033007241785526276, + "mean_feedback_forward_cosine": 0.99974972778751, + "mean_feedback_forward_relative_error": 0.02187272048585357, + "min_feedback_forward_cosine": 0.9994593858718872 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014160072170194369, + "instruction_rms": 0.0014160072167058446, + "raw_apical_rms": 0.07918109159981292, + "teaching_rms": 0.0014160072170194369, + "traffic_rms": 0.07916295968253205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.110519272326201e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.515495028598973e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014224182328554595, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.011447750072479248 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0330021008849144, + "mean_feedback_forward_cosine": 0.999749812387651, + "mean_feedback_forward_relative_error": 0.02186850042292668, + "min_feedback_forward_cosine": 0.9994595050811768 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013933700395346994, + "instruction_rms": 0.0013933700392033319, + "raw_apical_rms": 0.07918157824278696, + "teaching_rms": 0.0013933700395346994, + "traffic_rms": 0.07916338343285519 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.707011372483976e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510750225411907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01394252116038533, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.011034861965775489 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03299662470817566, + "mean_feedback_forward_cosine": 0.99974988929687, + "mean_feedback_forward_relative_error": 0.021864212313366514, + "min_feedback_forward_cosine": 0.9994596838951111 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001368839589429481, + "instruction_rms": 0.0013688395893179711, + "raw_apical_rms": 0.07917062061491366, + "teaching_rms": 0.001368839589429481, + "traffic_rms": 0.0791535977046907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.2251864234258392e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513117292228353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013996948203743328, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.01080519924364984 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03299080953001976, + "mean_feedback_forward_cosine": 0.999749977742472, + "mean_feedback_forward_relative_error": 0.021859597069241347, + "min_feedback_forward_cosine": 0.9994598627090454 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014290478374085786, + "instruction_rms": 0.0014290478379404262, + "raw_apical_rms": 0.07915993932216328, + "teaching_rms": 0.0014290478374085786, + "traffic_rms": 0.07914083382911073 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0593859656182758e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514032748636662e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013886441921894333, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.011204586127251387 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03298547491431236, + "mean_feedback_forward_cosine": 0.9997500815699177, + "mean_feedback_forward_relative_error": 0.02185536186481195, + "min_feedback_forward_cosine": 0.999459981918335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013227167268413406, + "instruction_rms": 0.001322716727390108, + "raw_apical_rms": 0.07913041551307097, + "teaching_rms": 0.0013227167268413406, + "traffic_rms": 0.07911367080930927 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0352396628832139e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5127181960960118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014147333314633512, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.010198485948964954 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032979924231767654, + "mean_feedback_forward_cosine": 0.9997501565564063, + "mean_feedback_forward_relative_error": 0.021850934071886922, + "min_feedback_forward_cosine": 0.9994602799415588 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001376275402272166, + "instruction_rms": 0.0013762754027576624, + "raw_apical_rms": 0.07909923910477713, + "teaching_rms": 0.001376275402272166, + "traffic_rms": 0.07908093860126976 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 7.992051109795284e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.511808489834985e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014044196766188915, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.010641923874206842 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03297405317425728, + "mean_feedback_forward_cosine": 0.9997502584611216, + "mean_feedback_forward_relative_error": 0.02184631022053861, + "min_feedback_forward_cosine": 0.9994604587554932 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014201690924522756, + "instruction_rms": 0.0014201690920126423, + "raw_apical_rms": 0.07914313714380547, + "teaching_rms": 0.0014201690924522756, + "traffic_rms": 0.07912430447505304 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.200623196647939e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5163667722108673e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01394759487130833, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.011176317230463029 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296904265880585, + "mean_feedback_forward_cosine": 0.9997503180657664, + "mean_feedback_forward_relative_error": 0.021842379107951157, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00125620757200252, + "instruction_rms": 0.0012562075712499428, + "raw_apical_rms": 0.0790734972576514, + "teaching_rms": 0.00125620757200252, + "traffic_rms": 0.07905877775397716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1262259391742191e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.515955062305274e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014052499863172833, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.009651345887109638 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968949526548386, + "mean_feedback_forward_cosine": 0.9997503238339578, + "mean_feedback_forward_relative_error": 0.02184228008733161, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012127444147503, + "instruction_rms": 0.0012127444148545786, + "raw_apical_rms": 0.07907800722754545, + "teaching_rms": 0.0012127444147503, + "traffic_rms": 0.07906396720153161 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0901062097445902e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5136365022952516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01390933908113494, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.009107687174677848 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296886384487152, + "mean_feedback_forward_cosine": 0.9997503296021493, + "mean_feedback_forward_relative_error": 0.021842192392796278, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012044832049351271, + "instruction_rms": 0.0012044832044063598, + "raw_apical_rms": 0.07907504564911902, + "teaching_rms": 0.0012044832049351271, + "traffic_rms": 0.07906055223895456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0774364511689782e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.517670431963739e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013905316794414391, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.009388247827142477 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296883404254913, + "mean_feedback_forward_cosine": 0.9997503411385321, + "mean_feedback_forward_relative_error": 0.021842103541618394, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011686305647989979, + "instruction_rms": 0.0011686305645142846, + "raw_apical_rms": 0.07907672729787814, + "teaching_rms": 0.0011686305647989979, + "traffic_rms": 0.07906340510438958 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.0281117823751629e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.518394430389037e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014070420830134678, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.008710122377723456 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296877071261406, + "mean_feedback_forward_cosine": 0.999750348829454, + "mean_feedback_forward_relative_error": 0.021842013608904615, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011258742416508797, + "instruction_rms": 0.0011258742418373587, + "raw_apical_rms": 0.0790740132440427, + "teaching_rms": 0.0011258742416508797, + "traffic_rms": 0.07906119572531516 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0154835905495929e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.517110430964284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013724314104478141, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.008479624182283879 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296871483325958, + "mean_feedback_forward_cosine": 0.9997503353703406, + "mean_feedback_forward_relative_error": 0.02184192481781206, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001143909574712528, + "instruction_rms": 0.001143909574256564, + "raw_apical_rms": 0.07907794714577587, + "teaching_rms": 0.001143909574712528, + "traffic_rms": 0.07906466208946192 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.976192589216225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5212801878121047e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014055417777558253, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.008150725717097521 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968681305646896, + "mean_feedback_forward_cosine": 0.999750348829454, + "mean_feedback_forward_relative_error": 0.021841845715478543, + "min_feedback_forward_cosine": 0.999460756778717 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001113586573641177, + "instruction_rms": 0.001113586574168122, + "raw_apical_rms": 0.07907284645007709, + "teaching_rms": 0.001113586573641177, + "traffic_rms": 0.07906011026335273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.142130585662926e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5168162832984197e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013956785087086336, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.008282279493808746 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296864777803421, + "mean_feedback_forward_cosine": 0.9997503392158016, + "mean_feedback_forward_relative_error": 0.02184175962822572, + "min_feedback_forward_cosine": 0.9994608163833618 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010981999131193143, + "instruction_rms": 0.00109819991333166, + "raw_apical_rms": 0.07907158907386407, + "teaching_rms": 0.0010981999131193143, + "traffic_rms": 0.07905950387920319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.484158880468385e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.51430887292063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013811410831241175, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.007810450859293341 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296860679984093, + "mean_feedback_forward_cosine": 0.9997503565203759, + "mean_feedback_forward_relative_error": 0.021841676274855292, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010293274383823628, + "instruction_rms": 0.0010293274385340119, + "raw_apical_rms": 0.07906876547630647, + "teaching_rms": 0.0010293274383823628, + "traffic_rms": 0.07905777306666822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0873407063627383e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.520211796338585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014243086572885448, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.007417631448805332 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296856954693794, + "mean_feedback_forward_cosine": 0.9997503507521844, + "mean_feedback_forward_relative_error": 0.021841591209053032, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010990896054554713, + "instruction_rms": 0.0010990896050859494, + "raw_apical_rms": 0.07907188351905532, + "teaching_rms": 0.0010990896054554713, + "traffic_rms": 0.07905936494461577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.216087448749022e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5128722676911396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01390453691567737, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.00799274047780782 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968513667583466, + "mean_feedback_forward_cosine": 0.9997503353703406, + "mean_feedback_forward_relative_error": 0.021841514780516585, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001039693161321965, + "instruction_rms": 0.0010396931615913152, + "raw_apical_rms": 0.07907147451871269, + "teaching_rms": 0.001039693161321965, + "traffic_rms": 0.07906007519289633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.175548143502965e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5159678167948397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013929542799831694, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.007195238195806742 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296850621700287, + "mean_feedback_forward_cosine": 0.9997503411385321, + "mean_feedback_forward_relative_error": 0.021841427010874593, + "min_feedback_forward_cosine": 0.9994604587554932 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010587673045355833, + "instruction_rms": 0.0010587673048441063, + "raw_apical_rms": 0.07906853031965344, + "teaching_rms": 0.0010587673045355833, + "traffic_rms": 0.07905707695607361 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0061324218213485e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5142331724915705e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013949775043342588, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.007522924195155501 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296847641468048, + "mean_feedback_forward_cosine": 0.9997503699794892, + "mean_feedback_forward_relative_error": 0.021841345940746607, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001032601818717189, + "instruction_rms": 0.0010326018188862908, + "raw_apical_rms": 0.07906675683980931, + "teaching_rms": 0.001032601818717189, + "traffic_rms": 0.07905548687416446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.145840351720835e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513182585962524e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013791341999313844, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.007077542001754046 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968468964099884, + "mean_feedback_forward_cosine": 0.9997503565203759, + "mean_feedback_forward_relative_error": 0.021841263383506768, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010075389793597076, + "instruction_rms": 0.001007538979640138, + "raw_apical_rms": 0.07905920973494185, + "teaching_rms": 0.0010075389793597076, + "traffic_rms": 0.07904918442559099 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.2415040373525699e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5149577056817836e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013963581488269363, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.007016531794071197 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968442887067795, + "mean_feedback_forward_cosine": 0.9997503276794187, + "mean_feedback_forward_relative_error": 0.021841186354117047, + "min_feedback_forward_cosine": 0.9994605183601379 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009794548225085466, + "instruction_rms": 0.000979454822212397, + "raw_apical_rms": 0.07906025538834051, + "teaching_rms": 0.0009794548225085466, + "traffic_rms": 0.07905011493779528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.708836756772143e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5159453615433813e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014014949148318256, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.0068274441438168285 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968390733003616, + "mean_feedback_forward_cosine": 0.9997503584431063, + "mean_feedback_forward_relative_error": 0.021841106290418293, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000954442312528871, + "instruction_rms": 0.000954442312153421, + "raw_apical_rms": 0.07905498202553352, + "teaching_rms": 0.000954442312528871, + "traffic_rms": 0.07904550296684917 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.484836655497929e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5164051989572564e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013849536754940474, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.006930758911669254 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296835720539093, + "mean_feedback_forward_cosine": 0.9997503411385321, + "mean_feedback_forward_relative_error": 0.021841023402709152, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010581227319160036, + "instruction_rms": 0.0010581227317756794, + "raw_apical_rms": 0.07905369779438841, + "teaching_rms": 0.0010581227319160036, + "traffic_rms": 0.07904225243721603 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1420457468602763e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5157561224054275e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014117948170150052, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.007310915174931288 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296829015016556, + "mean_feedback_forward_cosine": 0.9997503584431063, + "mean_feedback_forward_relative_error": 0.021840944976335572, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009231001149004738, + "instruction_rms": 0.0009231001146256697, + "raw_apical_rms": 0.07904741986531665, + "teaching_rms": 0.0009231001149004738, + "traffic_rms": 0.0790382252821156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.967943981337157e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.51251331573601e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01412589656533901, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0064019406197965145 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296823427081108, + "mean_feedback_forward_cosine": 0.9997503815158721, + "mean_feedback_forward_relative_error": 0.021840862343989072, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009712137624356089, + "instruction_rms": 0.0009712137624081436, + "raw_apical_rms": 0.07904598387235957, + "teaching_rms": 0.0009712137624356089, + "traffic_rms": 0.0790361511473722 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.2918070524392533e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510792366926753e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014148175272320715, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.006712962313741445 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968200743198395, + "mean_feedback_forward_cosine": 0.9997503565203759, + "mean_feedback_forward_relative_error": 0.021840780312495846, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009585568874126313, + "instruction_rms": 0.000958556887707422, + "raw_apical_rms": 0.07904796009636961, + "teaching_rms": 0.0009585568874126313, + "traffic_rms": 0.07903824963562338 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.406608825202056e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5158510592307154e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014014358810283886, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.006386181721538305 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296816349029541, + "mean_feedback_forward_cosine": 0.999750348829454, + "mean_feedback_forward_relative_error": 0.02184070112003434, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010190728852142073, + "instruction_rms": 0.001019072885635134, + "raw_apical_rms": 0.07904376617008083, + "teaching_rms": 0.0010190728852142073, + "traffic_rms": 0.07903246938581641 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 9.508214304787543e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5147979536538562e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014128765496529347, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.00735633303925395 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968126237392426, + "mean_feedback_forward_cosine": 0.9997503545976454, + "mean_feedback_forward_relative_error": 0.021840617376109286, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010597441145455456, + "instruction_rms": 0.0010597441143722257, + "raw_apical_rms": 0.07903477734099151, + "teaching_rms": 0.0010597441145455456, + "traffic_rms": 0.07902286833676171 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.194368916747408e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5194051803493045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014033613988649298, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.007420399015843868 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296808898448944, + "mean_feedback_forward_cosine": 0.9997503430612625, + "mean_feedback_forward_relative_error": 0.021840542705068664, + "min_feedback_forward_cosine": 0.9994608163833618 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009092463332496438, + "instruction_rms": 0.0009092463329607667, + "raw_apical_rms": 0.07902699670635355, + "teaching_rms": 0.0009092463332496438, + "traffic_rms": 0.07901763732489243 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1640194941264781e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5172593750667414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013780072240533117, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.006279832129329443 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296803683042526, + "mean_feedback_forward_cosine": 0.9997503603658369, + "mean_feedback_forward_relative_error": 0.0218404695211399, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009283214847513154, + "instruction_rms": 0.0009283214850865595, + "raw_apical_rms": 0.07902539066494889, + "teaching_rms": 0.0009283214847513154, + "traffic_rms": 0.0790161656110024 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0106430181622272e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510396423546384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013900489463075089, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.006302440658658743 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032968029379844666, + "mean_feedback_forward_cosine": 0.9997503680567588, + "mean_feedback_forward_relative_error": 0.021840394579715306, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009633431514219273, + "instruction_rms": 0.0009633431512511766, + "raw_apical_rms": 0.07902768305575265, + "teaching_rms": 0.0009633431514219273, + "traffic_rms": 0.07901792463611976 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0518120774349882e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5128072079863254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013987146702412094, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.006698597816675901 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296799957752228, + "mean_feedback_forward_cosine": 0.9997503565203759, + "mean_feedback_forward_relative_error": 0.021840316393683033, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010000299154175943, + "instruction_rms": 0.0010000299151706183, + "raw_apical_rms": 0.07902469276990336, + "teaching_rms": 0.0010000299154175943, + "traffic_rms": 0.0790144353089952 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 7.862864492835797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5112782806721006e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014643221676709678, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.006783464626297355 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967958599328995, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.021840241827791738, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008780621072531113, + "instruction_rms": 0.0008780621068969911, + "raw_apical_rms": 0.07902535642722684, + "teaching_rms": 0.0008780621072531113, + "traffic_rms": 0.07901718988832938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.293840008695042e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5190192119950504e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014337392750427867, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.006023441557586193 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296790271997452, + "mean_feedback_forward_cosine": 0.999750348829454, + "mean_feedback_forward_relative_error": 0.021840161869242307, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010084789308129687, + "instruction_rms": 0.00100847893088988, + "raw_apical_rms": 0.07901621026423537, + "teaching_rms": 0.0010084789308129687, + "traffic_rms": 0.07900551782606892 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1451479764446049e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5101286456537437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01404842447167463, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0068005267182737585 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296787291765213, + "mean_feedback_forward_cosine": 0.9997503661340282, + "mean_feedback_forward_relative_error": 0.02184008601151647, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009281412579568421, + "instruction_rms": 0.0009281412575446859, + "raw_apical_rms": 0.07901848716659302, + "teaching_rms": 0.0009281412579568421, + "traffic_rms": 0.07900935628414446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.991509841711377e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513563349317439e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014266434664338122, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0063689275822788475 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296784311532974, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.02184000653364966, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009401112224684706, + "instruction_rms": 0.0009401112219936799, + "raw_apical_rms": 0.07901911226798321, + "teaching_rms": 0.0009401112224684706, + "traffic_rms": 0.07900958266821878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 7.831042125872045e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5107561904213066e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013947600001987413, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.006227054849639535 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296780586242676, + "mean_feedback_forward_cosine": 0.9997503584431063, + "mean_feedback_forward_relative_error": 0.021839930045027888, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009164359653864836, + "instruction_rms": 0.000916435965763419, + "raw_apical_rms": 0.07901875888850046, + "teaching_rms": 0.0009164359653864836, + "traffic_rms": 0.07901024929324467 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.639287412710473e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5138816225905168e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014516120322245647, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.006150956730246544 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296780586242676, + "mean_feedback_forward_cosine": 0.9997503661340282, + "mean_feedback_forward_relative_error": 0.021839853661555435, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010130038828007153, + "instruction_rms": 0.0010130038828723808, + "raw_apical_rms": 0.07902319780907384, + "teaching_rms": 0.0010130038828007153, + "traffic_rms": 0.0790124095618252 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.684595170256216e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5105054896223538e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014464864722658278, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.006684562460407615 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296775743365288, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.02183977739825364, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009672630980360208, + "instruction_rms": 0.0009672630980452365, + "raw_apical_rms": 0.07902147399329391, + "teaching_rms": 0.0009672630980360208, + "traffic_rms": 0.07901109540736853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.772029502779333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5147833422659638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014153282180992266, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.006518044872879982 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296773135662079, + "mean_feedback_forward_cosine": 0.9997503603658369, + "mean_feedback_forward_relative_error": 0.021839702817341013, + "min_feedback_forward_cosine": 0.999460756778717 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008945342574852321, + "instruction_rms": 0.0008945342576079396, + "raw_apical_rms": 0.07902197896548262, + "teaching_rms": 0.0008945342574852321, + "traffic_rms": 0.07901360048996826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0292936281075526e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5145038787534723e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014322578794122146, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.006084084855392575 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0329676978290081, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.02183962811625773, + "min_feedback_forward_cosine": 0.9994605183601379 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008992423666483758, + "instruction_rms": 0.00089924236675753, + "raw_apical_rms": 0.07902933837563056, + "teaching_rms": 0.0008992423666483758, + "traffic_rms": 0.07902068264415925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0374198922136202e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5135901896982387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014059817482893236, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.006096628274694085 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967641949653625, + "mean_feedback_forward_cosine": 0.9997503642112978, + "mean_feedback_forward_relative_error": 0.02183955101176135, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009439295918184116, + "instruction_rms": 0.0009439295917520948, + "raw_apical_rms": 0.07903089944121158, + "teaching_rms": 0.0009439295918184116, + "traffic_rms": 0.07902142464199287 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 1.131690225975035e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513015308590563e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014110760511991617, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.006555102370977402 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296760842204094, + "mean_feedback_forward_cosine": 0.9997503661340282, + "mean_feedback_forward_relative_error": 0.02183947269053709, + "min_feedback_forward_cosine": 0.9994605779647827 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009616176574595979, + "instruction_rms": 0.0009616176575442036, + "raw_apical_rms": 0.07902639739093784, + "teaching_rms": 0.0009616176574595979, + "traffic_rms": 0.0790164520945508 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.79134915976465e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5130737708500898e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013927311754260987, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.006420740809813142 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296755254268646, + "mean_feedback_forward_cosine": 0.9997503680567588, + "mean_feedback_forward_relative_error": 0.0218393943542915, + "min_feedback_forward_cosine": 0.999460756778717 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009561128928233094, + "instruction_rms": 0.0009561128929356521, + "raw_apical_rms": 0.07901602955702663, + "teaching_rms": 0.0009561128928233094, + "traffic_rms": 0.07900628366509403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.298380294036178e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5198650270715774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014109272161080539, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.006626169956922531 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296753019094467, + "mean_feedback_forward_cosine": 0.9997503661340282, + "mean_feedback_forward_relative_error": 0.021839321245469393, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000840724984697107, + "instruction_rms": 0.0008407249842759102, + "raw_apical_rms": 0.07901127431004264, + "teaching_rms": 0.000840724984697107, + "traffic_rms": 0.07900356784083713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 1.0973425190741182e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510839297349668e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014017268517362923, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.005670979055911303 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296751528978348, + "mean_feedback_forward_cosine": 0.9997503719022197, + "mean_feedback_forward_relative_error": 0.02183924743064469, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009584272369043849, + "instruction_rms": 0.0009584272367005154, + "raw_apical_rms": 0.07900784757432908, + "teaching_rms": 0.0009584272369043849, + "traffic_rms": 0.0789979160255518 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485995054244995, + "maximum_absolute_post_projection_soma_slope": 1.1700515756274399e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514845728999968e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013975331223805388, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.006382069548889995 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296748921275139, + "mean_feedback_forward_cosine": 0.9997503545976454, + "mean_feedback_forward_relative_error": 0.02183917286475339, + "min_feedback_forward_cosine": 0.999460756778717 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009023088347941386, + "instruction_rms": 0.000902308834552656, + "raw_apical_rms": 0.07900471276416937, + "teaching_rms": 0.0009023088347941386, + "traffic_rms": 0.07899559234368503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.519089871901087e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.51414561701434e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014107934734276593, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.00600534839913249 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967474311590195, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.02183909953061131, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009122620665221869, + "instruction_rms": 0.0009122620665402818, + "raw_apical_rms": 0.07900576117634668, + "teaching_rms": 0.0009122620665221869, + "traffic_rms": 0.07899680517053587 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485990583896637, + "maximum_absolute_post_projection_soma_slope": 8.163934950289331e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5162347607435555e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01422119973126778, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.006067476005852223 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967422157526016, + "mean_feedback_forward_cosine": 0.9997503680567588, + "mean_feedback_forward_relative_error": 0.021839025550551953, + "min_feedback_forward_cosine": 0.999460756778717 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008872481055347488, + "instruction_rms": 0.0008872481059716014, + "raw_apical_rms": 0.07900496911237236, + "teaching_rms": 0.0008872481055347488, + "traffic_rms": 0.07899654099239764 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.099937776416482e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.517637085567607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013890023698339776, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.005863672554884106 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296740725636482, + "mean_feedback_forward_cosine": 0.9997503719022197, + "mean_feedback_forward_relative_error": 0.021838953027561787, + "min_feedback_forward_cosine": 0.9994608163833618 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009156327037024176, + "instruction_rms": 0.0009156327038337892, + "raw_apical_rms": 0.0790039321496955, + "teaching_rms": 0.0009156327037024176, + "traffic_rms": 0.07899460412410766 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 9.450116067455383e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5189542867007598e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014710242127299952, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.00601576112754643 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967373728752136, + "mean_feedback_forward_cosine": 0.9997503642112978, + "mean_feedback_forward_relative_error": 0.02183888137580887, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009124955672854133, + "instruction_rms": 0.0009124955673140389, + "raw_apical_rms": 0.07900084201321039, + "teaching_rms": 0.0009124955672854133, + "traffic_rms": 0.07899186077277014 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.168264773809824e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5140803261038174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014547152937846046, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.005961679786294699 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296734765172005, + "mean_feedback_forward_cosine": 0.9997503699794892, + "mean_feedback_forward_relative_error": 0.021838807365706852, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009445591740789052, + "instruction_rms": 0.0009445591741396572, + "raw_apical_rms": 0.07899786446561065, + "teaching_rms": 0.0009445591740789052, + "traffic_rms": 0.07898848974266093 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1056327053893256e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.516956032154085e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014368800921630846, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.006252165116071701 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296731784939766, + "mean_feedback_forward_cosine": 0.9997503622885673, + "mean_feedback_forward_relative_error": 0.021838735338420637, + "min_feedback_forward_cosine": 0.9994606375694275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009108061973677489, + "instruction_rms": 0.0009108061971431601, + "raw_apical_rms": 0.07899758507212541, + "teaching_rms": 0.0009108061973677489, + "traffic_rms": 0.07898874183738822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.3652264385655144e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5132383642144577e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01409900083149227, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.006082479944527149 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967302948236465, + "mean_feedback_forward_cosine": 0.9997503719022197, + "mean_feedback_forward_relative_error": 0.021838656912047054, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009383804513924885, + "instruction_rms": 0.0009383804510292355, + "raw_apical_rms": 0.07899267105418258, + "teaching_rms": 0.0009383804513924885, + "traffic_rms": 0.07898357952351737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.1492713269944943e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5191171640399516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014221252394748479, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.006195355851203203 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.032967254519462585, + "mean_feedback_forward_cosine": 0.9997503815158721, + "mean_feedback_forward_relative_error": 0.02183858228607043, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008929500694685327, + "instruction_rms": 0.0008929500696270735, + "raw_apical_rms": 0.07899459633576932, + "teaching_rms": 0.0008929500694685327, + "traffic_rms": 0.07898567271487614 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485992074012756, + "maximum_absolute_post_projection_soma_slope": 8.929259109891063e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5141642521591798e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013872893435305468, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.005830887074619532 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0329672209918499, + "mean_feedback_forward_cosine": 0.9997503565203759, + "mean_feedback_forward_relative_error": 0.02183850734464584, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008716643037513351, + "instruction_rms": 0.000871664303017085, + "raw_apical_rms": 0.07899832843137548, + "teaching_rms": 0.0008716643037513351, + "traffic_rms": 0.07898993315542388 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 1.0306368380952335e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5132546889416105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014071420884752703, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.005806158569604159 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03296716883778572, + "mean_feedback_forward_cosine": 0.9997503872840635, + "mean_feedback_forward_relative_error": 0.021838433800205108, + "min_feedback_forward_cosine": 0.9994606971740723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009030733433882777, + "instruction_rms": 0.0009030733431554299, + "raw_apical_rms": 0.07899734028753069, + "teaching_rms": 0.0009030733433882777, + "traffic_rms": 0.07898853607377436 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.16485993564128876, + "maximum_absolute_post_projection_soma_slope": 9.832930913944438e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5207141787821894e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014025337666491038, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.005745448650866747 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9245, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3702026550292969, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2529336320, + "peak_memory_reserved_bytes": 3821010944, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 4.7853433216005214e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 3.677267201851464e-08, + "min_residual_soma_slope": -4.7853433216005214e-08, + "mse": 5.5042523997636e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.121279244416156e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.121279244416156e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 1.1455144882202148, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.04877066612243652, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 14949.984804391861, + "train_wall_s": 14945.887037992477 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.022805234417319298, + 0.016488894820213318, + 0.010465242899954319, + 0.009960980154573917, + 0.0061687263660132885, + 0.0069997189566493034, + 0.0042089177295565605, + 0.0052098724991083145, + 0.0032430493738502264, + 0.004236941691488028, + 0.0026123940479010344, + 0.005416125524789095, + 0.0038156001828610897, + 0.004767284262925386, + 0.0030251212883740664, + 0.004124426748603582, + 0.0025168154388666153, + 0.003442907240241766, + 0.0021780587267130613, + 0.0030431649647653103, + 0.0018720370717346668, + 0.004256135784089565, + 0.0029823817312717438, + 0.003729435382410884, + 0.0026436245534569025, + 0.003457642626017332, + 0.002387387678027153, + 0.0029531591571867466, + 0.0022456953302025795, + 0.002684617880731821, + 0.002184882527217269 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.000000476837158, + 3.999999523162842, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 3.999999523162842, + 3.999999761581421, + 3.999999761581421, + 4.0, + 3.999999761581421, + 3.999999523162842, + 4.0, + 4.0, + 3.999999761581421, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.12281870096921921, + 0.0890195369720459, + 0.037514179944992065, + 0.05279386788606644, + 0.016393808647990227, + 0.03781002387404442, + 0.009272366762161255, + 0.028348324820399284, + 0.006517303641885519, + 0.022264672443270683, + 0.004724749829620123, + 0.028529828414320946, + 0.008454201743006706, + 0.0255922582000494, + 0.006007933523505926, + 0.022076090797781944, + 0.004486671183258295, + 0.018180686980485916, + 0.0035881344228982925, + 0.01652967743575573, + 0.0028566550463438034, + 0.022440683096647263, + 0.005847430322319269, + 0.019685085862874985, + 0.004685010761022568, + 0.018438123166561127, + 0.003933165688067675, + 0.015699174255132675, + 0.0034667328000068665, + 0.014064486138522625, + 0.003166335867717862 + ], + "unscaled_traffic_rms": [ + 0.7427284121513367, + 0.7409113049507141, + 1.1158705949783325, + 0.7547073364257812, + 1.5051356554031372, + 0.7405146360397339, + 1.8156821727752686, + 0.7351224422454834, + 1.9904239177703857, + 0.7611954212188721, + 2.211667537689209, + 0.7593632340431213, + 1.8053035736083984, + 0.7451134920120239, + 2.0140843391418457, + 0.7473110556602478, + 2.2438154220581055, + 0.7574867010116577, + 2.428068161010742, + 0.7364124059677124, + 2.6212995052337646, + 0.7586463689804077, + 2.0401315689086914, + 0.7578194737434387, + 2.257091522216797, + 0.7501072883605957, + 2.427955389022827, + 0.7524368166923523, + 2.591137409210205, + 0.7635167837142944, + 2.7601399421691895 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 684206618894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 684202240000000, + "local_weight_correlation_macs": 688625920000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 688625920000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 72745580756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 2745665106100224 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json new file mode 100644 index 0000000..6ec049d --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json @@ -0,0 +1,7293 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 10, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 10, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5010, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9994682272275289, + "feedback_forward_cosine": [ + 0.9994404911994934, + 0.9993954300880432, + 0.9997115135192871, + 0.9996904730796814, + 0.999552845954895, + 0.9995489120483398, + 0.9995783567428589, + 0.9995408058166504, + 0.9996228218078613, + 0.999614953994751, + 0.9996095299720764, + 0.999545693397522, + 0.9996140003204346, + 0.9995242357254028, + 0.999771773815155, + 0.9996898174285889, + 0.9997661709785461, + 0.9996163845062256, + 0.9997018575668335, + 0.9997928142547607, + 0.9996846914291382, + 0.999636709690094, + 0.9997365474700928, + 0.9996662139892578, + 0.9997066259384155, + 0.9996163845062256, + 0.9997081756591797, + 0.9996177554130554, + 0.999658465385437, + 0.9995304346084595, + 0.9996687173843384, + 0.9995114803314209, + 0.9995819330215454, + 0.9993857145309448, + 0.9994992017745972, + 0.9992763996124268, + 0.9997461438179016, + 0.9998161196708679, + 0.9997888803482056, + 0.9997533559799194, + 0.9997977018356323, + 0.9996863603591919, + 0.9997535347938538, + 0.99958735704422, + 0.999610185623169, + 0.9993331432342529, + 0.999364972114563, + 0.9990265369415283, + 0.9985905885696411, + 0.9981398582458496, + 0.9982001781463623, + 0.9977648258209229, + 0.9981805086135864, + 0.9975036978721619, + 0.9999802708625793 + ], + "feedback_forward_norm_ratio": [ + 0.9990659356117249, + 1.0005857944488525, + 0.9993277192115784, + 1.0003453493118286, + 0.9995670318603516, + 1.000895380973816, + 0.9981652498245239, + 1.0008583068847656, + 0.9991608262062073, + 1.001194953918457, + 0.9992017149925232, + 1.0010004043579102, + 0.9991812705993652, + 1.002564549446106, + 0.9990981817245483, + 1.000566005706787, + 0.9995611906051636, + 1.0014108419418335, + 0.9984918236732483, + 1.0005040168762207, + 0.9991447925567627, + 1.0008258819580078, + 0.9996499419212341, + 1.0017189979553223, + 0.9990311861038208, + 1.002008080482483, + 0.9997596740722656, + 1.001738429069519, + 0.9991170167922974, + 1.0017306804656982, + 0.9995566010475159, + 1.0015692710876465, + 0.9979696869850159, + 1.0022270679473877, + 0.9989573359489441, + 1.0032389163970947, + 0.9992907047271729, + 1.0006147623062134, + 0.9997590780258179, + 1.0006561279296875, + 0.9998376369476318, + 1.0011348724365234, + 0.9995772838592529, + 1.0010188817977905, + 0.9992583990097046, + 1.0021554231643677, + 0.99852454662323, + 1.00221586227417, + 0.9974808096885681, + 1.0032844543457031, + 0.9973658323287964, + 1.004159688949585, + 0.9972424507141113, + 1.0035505294799805, + 1.001733660697937 + ], + "feedback_forward_relative_error": [ + 0.03344787284731865, + 0.03478962555527687, + 0.0240226648747921, + 0.024886533617973328, + 0.02990039624273777, + 0.03006056882441044, + 0.02907276526093483, + 0.030331334099173546, + 0.027464091777801514, + 0.0277912225574255, + 0.027941882610321045, + 0.0301781315356493, + 0.027781657874584198, + 0.03099004551768303, + 0.021371891722083092, + 0.02491869404911995, + 0.02162111923098564, + 0.027751967310905457, + 0.02444976009428501, + 0.020368635654449463, + 0.025114303454756737, + 0.02697877772152424, + 0.02295496314764023, + 0.02591652236878872, + 0.02422996051609516, + 0.02780137211084366, + 0.02415524236857891, + 0.027729131281375885, + 0.026138020679354668, + 0.030723802745342255, + 0.025735855102539062, + 0.031318821012973785, + 0.02895987033843994, + 0.03516266867518425, + 0.03164835274219513, + 0.03823957219719887, + 0.022533871233463287, + 0.019190732389688492, + 0.02054876834154129, + 0.02222895808517933, + 0.020110785961151123, + 0.025087302550673485, + 0.022203581407666206, + 0.02875879406929016, + 0.02792232856154442, + 0.03662370890378952, + 0.03564311936497688, + 0.044228386133909225, + 0.05308665335178375, + 0.06118026003241539, + 0.05997497960925102, + 0.06712844967842102, + 0.06030327081680298, + 0.07087264955043793, + 0.006518933922052383 + ], + "innovation_negative_gradient_cosine": [ + 0.9995369911193848, + 0.9993072748184204, + 0.9995357990264893, + 0.9994770884513855, + 0.9995355010032654, + 0.9994328618049622, + 0.9995237588882446, + 0.9993277788162231, + 0.9995232820510864, + 0.9993844032287598, + 0.9995205402374268, + 0.9993714094161987, + 0.9995244741439819, + 0.999367356300354, + 0.9995344877243042, + 0.9995198249816895, + 0.9995425939559937, + 0.9994626641273499, + 0.9995633959770203, + 0.9995496273040771, + 0.9995813369750977, + 0.9994677305221558, + 0.999589741230011, + 0.9994921088218689, + 0.9996006488800049, + 0.9994814991950989, + 0.9996095895767212, + 0.9995163679122925, + 0.9996070861816406, + 0.999485433101654, + 0.9996163845062256, + 0.9995322227478027, + 0.9996259212493896, + 0.9994911551475525, + 0.9996343851089478, + 0.9994434714317322, + 0.9996479749679565, + 0.9996869564056396, + 0.999660313129425, + 0.9996445178985596, + 0.9996895790100098, + 0.9996963143348694, + 0.9996941089630127, + 0.9997339844703674, + 0.9996970891952515, + 0.9997127056121826, + 0.9996949434280396, + 0.9997172355651855, + 0.9995944499969482, + 0.9993976950645447, + 0.9996228814125061, + 0.9996196627616882, + 0.9996307492256165, + 0.9996695518493652, + 0.6467665433883667 + ], + "instruction_negative_gradient_cosine": [ + 0.9995369911193848, + 0.9993072748184204, + 0.9995357990264893, + 0.9994770288467407, + 0.9995355010032654, + 0.9994328618049622, + 0.9995237588882446, + 0.9993277788162231, + 0.9995232820510864, + 0.9993844628334045, + 0.9995205402374268, + 0.9993714094161987, + 0.9995244741439819, + 0.9993672370910645, + 0.9995344877243042, + 0.9995198249816895, + 0.9995425939559937, + 0.9994626641273499, + 0.999563455581665, + 0.9995495676994324, + 0.9995813369750977, + 0.9994677305221558, + 0.999589741230011, + 0.9994921088218689, + 0.9996007084846497, + 0.9994814991950989, + 0.9996095895767212, + 0.9995163679122925, + 0.9996070861816406, + 0.9994853734970093, + 0.9996163845062256, + 0.9995322227478027, + 0.9996258616447449, + 0.9994912147521973, + 0.999634325504303, + 0.9994434714317322, + 0.9996479749679565, + 0.9996869564056396, + 0.999660313129425, + 0.9996445178985596, + 0.9996895790100098, + 0.9996963143348694, + 0.9996941089630127, + 0.9997339844703674, + 0.9996970891952515, + 0.9997127056121826, + 0.9996948838233948, + 0.9997172355651855, + 0.9995944499969482, + 0.9993976950645447, + 0.9996228814125061, + 0.9996196031570435, + 0.9996308088302612, + 0.9996695518493652, + 0.6637846231460571 + ], + "matched_negative_gradient_cosine": [ + 0.007951020263135433, + 0.016651220619678497, + 0.00022909475956112146, + 0.0007789621595293283, + 0.0020209504291415215, + -0.003988915588706732, + 0.007779962383210659, + 0.0169359240680933, + 0.006374146323651075, + -0.019619127735495567, + 0.015027269721031189, + -0.0005737957544624805, + 0.012350887060165405, + 0.01950559765100479, + 0.009089230559766293, + 0.019061442464590073, + 0.010499964468181133, + 0.028676047921180725, + -0.00823073647916317, + 0.05451694130897522, + 0.00878145918250084, + 0.014558956027030945, + 0.014895726926624775, + 0.03158414363861084, + 0.03653767704963684, + 0.1080160140991211, + 0.04275117814540863, + 0.0927334874868393, + 0.033844441175460815, + 0.07671237736940384, + 0.04065904766321182, + 0.058761682361364365, + 0.052511297166347504, + 0.09380972385406494, + 0.054168861359357834, + 0.09256035089492798, + 0.05772949010133743, + 0.006526164244860411, + 0.04461504891514778, + 0.16173142194747925, + 0.019345808774232864, + 0.10187895596027374, + 0.024881921708583832, + 0.1399366557598114, + 0.019930541515350342, + 0.04394404590129852, + 0.040142178535461426, + 0.02651257999241352, + 0.056152936071157455, + 0.029059287160634995, + 0.02733425609767437, + 0.007577531039714813, + 0.026050828397274017, + 0.02375444583594799, + 0.06823229044675827 + ], + "max_norm_match_direction_error": 0.14550673961639404, + "max_norm_match_relative_error": 1.673303842153473e-07, + "mean_feedback_forward_cosine": 0.9994443373246626, + "mean_feedback_forward_relative_error": 0.031019884212450548, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.21679285168647766, + "max_absolute_post_projection_soma_slope": 2.920860175947837e-08, + "max_absolute_pre_projection_soma_slope": 0.21679285168647766, + "max_positive_post_projection_soma_slope": 2.920860175947837e-08, + "min_post_projection_soma_slope": -2.2999914861543402e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.3419398701970128e-08, + "pre_projection_traffic_rms_ratio": 0.005726907156141328 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.005726907156090022, + "raw_negative_gradient_cosine": [ + 0.007951020263135433, + 0.016651220619678497, + 0.00022909522522240877, + 0.0007789625087752938, + 0.0020209504291415215, + -0.0039889151230454445, + 0.007779961451888084, + 0.0169359240680933, + 0.006374146789312363, + -0.019619125872850418, + 0.015027269721031189, + -0.0005737952888011932, + 0.01235088612884283, + 0.01950559765100479, + 0.009089229628443718, + 0.019061442464590073, + 0.010499965399503708, + 0.028676047921180725, + -0.00823073647916317, + 0.05451694130897522, + 0.00878145918250084, + 0.01455895695835352, + 0.014895725063979626, + 0.03158414363861084, + 0.03653767704963684, + 0.1080160140991211, + 0.04275117814540863, + 0.0927334874868393, + 0.033844441175460815, + 0.07671236991882324, + 0.04065905138850212, + 0.058761678636074066, + 0.052511293441057205, + 0.09380972385406494, + 0.054168861359357834, + 0.09256035089492798, + 0.05772949755191803, + 0.00652616610750556, + 0.04461504891514778, + 0.16173143684864044, + 0.019345812499523163, + 0.10187895596027374, + 0.024881921708583832, + 0.1399366557598114, + 0.019930541515350342, + 0.04394404590129852, + 0.040142182260751724, + 0.02651258185505867, + 0.05615293234586716, + 0.029059283435344696, + 0.02733425796031952, + 0.007577531039714813, + 0.026050828397274017, + 0.02375444769859314, + 0.0681997612118721 + ], + "teaching_negative_gradient_cosine": [ + 0.9995369911193848, + 0.9993072748184204, + 0.9995357990264893, + 0.9994770884513855, + 0.9995355010032654, + 0.9994328618049622, + 0.9995237588882446, + 0.9993277788162231, + 0.9995232820510864, + 0.9993844032287598, + 0.9995205402374268, + 0.9993714094161987, + 0.9995244741439819, + 0.999367356300354, + 0.9995344877243042, + 0.9995198249816895, + 0.9995425939559937, + 0.9994626641273499, + 0.9995633959770203, + 0.9995496273040771, + 0.9995813369750977, + 0.9994677305221558, + 0.999589741230011, + 0.9994921088218689, + 0.9996006488800049, + 0.9994814991950989, + 0.9996095895767212, + 0.9995163679122925, + 0.9996070861816406, + 0.999485433101654, + 0.9996163845062256, + 0.9995322227478027, + 0.9996259212493896, + 0.9994911551475525, + 0.9996343851089478, + 0.9994434714317322, + 0.9996479749679565, + 0.9996869564056396, + 0.999660313129425, + 0.9996445178985596, + 0.9996895790100098, + 0.9996963143348694, + 0.9996941089630127, + 0.9997339844703674, + 0.9996970891952515, + 0.9997127056121826, + 0.9996949434280396, + 0.9997172355651855, + 0.9995944499969482, + 0.9993976950645447, + 0.9996228814125061, + 0.9996196627616882, + 0.9996307492256165, + 0.9996695518493652, + 0.6467665433883667 + ], + "traffic_instruction_rms_ratio": 166.47478673073695, + "used_negative_gradient_cosine": [ + 0.9995369911193848, + 0.9993072748184204, + 0.9995357990264893, + 0.9994770884513855, + 0.9995355010032654, + 0.9994328618049622, + 0.9995237588882446, + 0.9993277788162231, + 0.9995232820510864, + 0.9993844032287598, + 0.9995205402374268, + 0.9993714094161987, + 0.9995244741439819, + 0.999367356300354, + 0.9995344877243042, + 0.9995198249816895, + 0.9995425939559937, + 0.9994626641273499, + 0.9995633959770203, + 0.9995496273040771, + 0.9995813369750977, + 0.9994677305221558, + 0.999589741230011, + 0.9994921088218689, + 0.9996006488800049, + 0.9994814991950989, + 0.9996095895767212, + 0.9995163679122925, + 0.9996070861816406, + 0.999485433101654, + 0.9996163845062256, + 0.9995322227478027, + 0.9996259212493896, + 0.9994911551475525, + 0.9996343851089478, + 0.9994434714317322, + 0.9996479749679565, + 0.9996869564056396, + 0.999660313129425, + 0.9996445178985596, + 0.9996895790100098, + 0.9996963143348694, + 0.9996941089630127, + 0.9997339844703674, + 0.9996970891952515, + 0.9997127056121826, + 0.9996949434280396, + 0.9997172355651855, + 0.9995944499969482, + 0.9993976950645447, + 0.9996228814125061, + 0.9996196627616882, + 0.9996307492256165, + 0.9996695518493652, + 0.6467665433883667 + ], + "wall_s": 0.3682084083557129 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.369456171989441, + "mean_feedback_forward_cosine": 0.2689082251353697, + "mean_feedback_forward_relative_error": 1.1972714272412386, + "min_feedback_forward_cosine": 0.06917765736579895 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0014078608180658027, + "instruction_rms": 0.001407860820661047, + "raw_apical_rms": 0.07253148805079876, + "teaching_rms": 0.0014078608180658027, + "traffic_rms": 0.07242028298479802 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.183836459034865e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.0222815594380756e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008607884238328454, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.5390182747650147 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3501542806625366, + "mean_feedback_forward_cosine": 0.2927861137823625, + "mean_feedback_forward_relative_error": 1.177639068256725, + "min_feedback_forward_cosine": 0.0984971821308136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00124677806096627, + "instruction_rms": 0.001246778060357656, + "raw_apical_rms": 0.07905536325674647, + "teaching_rms": 0.00124677806096627, + "traffic_rms": 0.07904517099797034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.801659134860529e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.813901610059566e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009354999126080822, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.825491848373413 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3406648635864258, + "mean_feedback_forward_cosine": 0.33759623102166436, + "mean_feedback_forward_relative_error": 1.1397242833267558, + "min_feedback_forward_cosine": 0.10546259582042694 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0017762302135308115, + "instruction_rms": 0.0017762302137874684, + "raw_apical_rms": 0.09286173665831304, + "teaching_rms": 0.0017762302135308115, + "traffic_rms": 0.09284445870091278 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1403717081748255e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7028709890868054e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008458986768579237, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.5229721897888184 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3353369235992432, + "mean_feedback_forward_cosine": 0.38439204462549903, + "mean_feedback_forward_relative_error": 1.0989309056238694, + "min_feedback_forward_cosine": 0.11290597170591354 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002144352366856395, + "instruction_rms": 0.002144352366926018, + "raw_apical_rms": 0.10642397751297464, + "teaching_rms": 0.002144352366856395, + "traffic_rms": 0.10640207164306367 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.766463738789753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.607272010383771e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007169416539074072, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.2386383350372314 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3303279876708984, + "mean_feedback_forward_cosine": 0.42887628769332714, + "mean_feedback_forward_relative_error": 1.058563599803231, + "min_feedback_forward_cosine": 0.11963705718517303 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0024083291066470266, + "instruction_rms": 0.002408329106783188, + "raw_apical_rms": 0.1160738627277121, + "teaching_rms": 0.0024083291066470266, + "traffic_rms": 0.11604860936263153 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.463994305429878e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5641748357825903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007335635853544253, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.0158939805603027 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3247514963150024, + "mean_feedback_forward_cosine": 0.468974922191013, + "mean_feedback_forward_relative_error": 1.0200427906079725, + "min_feedback_forward_cosine": 0.127312570810318 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025600956808246975, + "instruction_rms": 0.0025600956817007758, + "raw_apical_rms": 0.12247912124202827, + "teaching_rms": 0.0025600956808246975, + "traffic_rms": 0.12245209271991864 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 5.8787094303625054e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5591618086320432e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008561494907804461, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8577664368057251 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3185664415359497, + "mean_feedback_forward_cosine": 0.5038931838490747, + "mean_feedback_forward_relative_error": 0.9849336450750178, + "min_feedback_forward_cosine": 0.13563823699951172 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025940226788831447, + "instruction_rms": 0.002594022678948535, + "raw_apical_rms": 0.12895238646516094, + "teaching_rms": 0.0025940226788831447, + "traffic_rms": 0.12892603267008163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.997678525522133e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5463127291706684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008545097883428839, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7557037147521972 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.312221646308899, + "mean_feedback_forward_cosine": 0.53461450798945, + "mean_feedback_forward_relative_error": 0.9526763005690141, + "min_feedback_forward_cosine": 0.1440867930650711 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026261802167477115, + "instruction_rms": 0.0026261802163296158, + "raw_apical_rms": 0.13317921947134523, + "teaching_rms": 0.0026261802167477115, + "traffic_rms": 0.1331530843718966 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.0903553970820212e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529052634860394e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00911506880900851, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6905287527656555 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3053839206695557, + "mean_feedback_forward_cosine": 0.5630068250677802, + "mean_feedback_forward_relative_error": 0.9216304621913216, + "min_feedback_forward_cosine": 0.15337881445884705 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026532173367286287, + "instruction_rms": 0.002653217336847133, + "raw_apical_rms": 0.13781177115967047, + "teaching_rms": 0.0026532173367286287, + "traffic_rms": 0.13778599137145367 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 9.864122318958835e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5253319297284716e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00903966909057363, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6335982405853271 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2983505725860596, + "mean_feedback_forward_cosine": 0.5881762472066012, + "mean_feedback_forward_relative_error": 0.8930000028826973, + "min_feedback_forward_cosine": 0.16186240315437317 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027004392122214776, + "instruction_rms": 0.0027004392135709186, + "raw_apical_rms": 0.1409414882365129, + "teaching_rms": 0.0027004392122214776, + "traffic_rms": 0.14091533321399727 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.604141837371571e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5159799066847384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008300611858488472, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5807088320827484 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.289905309677124, + "mean_feedback_forward_cosine": 0.611640171029351, + "mean_feedback_forward_relative_error": 0.8652345949953253, + "min_feedback_forward_cosine": 0.17215199768543243 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002716556502013636, + "instruction_rms": 0.002716556501728653, + "raw_apical_rms": 0.1438973105166632, + "teaching_rms": 0.002716556502013636, + "traffic_rms": 0.1438713914837601 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 9.660015365398067e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5125790295178934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008332734171125573, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5490889276123047 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2804360389709473, + "mean_feedback_forward_cosine": 0.6332393632693725, + "mean_feedback_forward_relative_error": 0.8388771290128881, + "min_feedback_forward_cosine": 0.1850309669971466 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027347042332565657, + "instruction_rms": 0.0027347042333227137, + "raw_apical_rms": 0.145752187024359, + "teaching_rms": 0.0027347042332565657, + "traffic_rms": 0.14572623077321176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.2038564989325096e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5105251607952445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008174152878091092, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5216856644821167 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2710715532302856, + "mean_feedback_forward_cosine": 0.6532921455123207, + "mean_feedback_forward_relative_error": 0.8137308101762425, + "min_feedback_forward_cosine": 0.1969955563545227 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027213841771664385, + "instruction_rms": 0.0027213841770934604, + "raw_apical_rms": 0.1482095211465228, + "teaching_rms": 0.0027213841771664385, + "traffic_rms": 0.1481842355957605 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.45189455592299e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.526094867671055e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008321533278653932, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.49639288541793825 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2606836557388306, + "mean_feedback_forward_cosine": 0.6722355338660153, + "mean_feedback_forward_relative_error": 0.7892406214367259, + "min_feedback_forward_cosine": 0.21050788462162018 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002706110810581065, + "instruction_rms": 0.0027061108111700195, + "raw_apical_rms": 0.1498301910981681, + "teaching_rms": 0.002706110810581065, + "traffic_rms": 0.1498054236433129 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.23983947132001e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.504863395961472e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008117705535747154, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.46903689533233645 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2496236562728882, + "mean_feedback_forward_cosine": 0.6898900433020159, + "mean_feedback_forward_relative_error": 0.7656778820536353, + "min_feedback_forward_cosine": 0.2246866524219513 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00268414078045173, + "instruction_rms": 0.002684140779454647, + "raw_apical_rms": 0.1510483561881407, + "teaching_rms": 0.00268414078045173, + "traffic_rms": 0.151024166507925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 1.180244240117645e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5135116606682998e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008063122030842956, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4508209700489044 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2381395101547241, + "mean_feedback_forward_cosine": 0.706576835567301, + "mean_feedback_forward_relative_error": 0.7427569503133947, + "min_feedback_forward_cosine": 0.24030257761478424 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027269388715657687, + "instruction_rms": 0.0027269388713653925, + "raw_apical_rms": 0.15219185098940602, + "teaching_rms": 0.0027269388715657687, + "traffic_rms": 0.1521671288481907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.0166957764567996e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5060952700744226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0077939087266589505, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.42989219255447386 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2247095108032227, + "mean_feedback_forward_cosine": 0.7227096731012518, + "mean_feedback_forward_relative_error": 0.7199545334685933, + "min_feedback_forward_cosine": 0.25628435611724854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027349959697728905, + "instruction_rms": 0.002734995970364187, + "raw_apical_rms": 0.15284853997319411, + "teaching_rms": 0.0027349959697728905, + "traffic_rms": 0.15282369517765512 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.373790822384763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5198463096368874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007536574376778029, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.4131422837734222 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2106099128723145, + "mean_feedback_forward_cosine": 0.7377519146962599, + "mean_feedback_forward_relative_error": 0.6982379024679011, + "min_feedback_forward_cosine": 0.27311837673187256 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002738938882674984, + "instruction_rms": 0.002738938882131211, + "raw_apical_rms": 0.1547192768412384, + "teaching_rms": 0.002738938882674984, + "traffic_rms": 0.15469464677270342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.65467325688951e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509952813512533e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007475800837032979, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.395740670375824 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1949330568313599, + "mean_feedback_forward_cosine": 0.7525744221427224, + "mean_feedback_forward_relative_error": 0.6761982630599629, + "min_feedback_forward_cosine": 0.29287850856781006 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002789933881637465, + "instruction_rms": 0.002789933882418844, + "raw_apical_rms": 0.15529511365498264, + "teaching_rms": 0.002789933881637465, + "traffic_rms": 0.15526965106451546 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.554450576208183e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5095342040844068e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007504243891270323, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3899913565635681 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1793348789215088, + "mean_feedback_forward_cosine": 0.7664710158651525, + "mean_feedback_forward_relative_error": 0.654895176670768, + "min_feedback_forward_cosine": 0.3113252520561218 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028157399067542995, + "instruction_rms": 0.0028157399068014966, + "raw_apical_rms": 0.1557017757088369, + "teaching_rms": 0.0028157399067542995, + "traffic_rms": 0.15567592398314006 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.2300030505230097e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5141737887963763e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007724938687399083, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3716805156517029 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1615961790084839, + "mean_feedback_forward_cosine": 0.7794803326780145, + "mean_feedback_forward_relative_error": 0.6345552338795228, + "min_feedback_forward_cosine": 0.3317318558692932 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027772103042668236, + "instruction_rms": 0.0027772103042611046, + "raw_apical_rms": 0.15703431703397072, + "teaching_rms": 0.0027772103042668236, + "traffic_rms": 0.15700935973156757 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.0139478945347946e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5200974792610012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007565728388337472, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3565149616909027 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1436591148376465, + "mean_feedback_forward_cosine": 0.7917030274868011, + "mean_feedback_forward_relative_error": 0.6147973905910146, + "min_feedback_forward_cosine": 0.3503715693950653 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002740800465849123, + "instruction_rms": 0.0027408004651419916, + "raw_apical_rms": 0.15884990233234111, + "teaching_rms": 0.002740800465849123, + "traffic_rms": 0.15882587576190477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 8.08871476465356e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.515649550919636e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007051738210346881, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.34919896123886107 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1289823055267334, + "mean_feedback_forward_cosine": 0.8035282308405096, + "mean_feedback_forward_relative_error": 0.5955030484633013, + "min_feedback_forward_cosine": 0.37110692262649536 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002720008624748741, + "instruction_rms": 0.0027200086257573902, + "raw_apical_rms": 0.1591613560909939, + "teaching_rms": 0.002720008624748741, + "traffic_rms": 0.1591376994669512 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 5.961846483160116e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507685907145174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006787212674603808, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3399496278190613 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1149834394454956, + "mean_feedback_forward_cosine": 0.8148032464764335, + "mean_feedback_forward_relative_error": 0.5765755279497666, + "min_feedback_forward_cosine": 0.3912418484687805 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027380940860072885, + "instruction_rms": 0.002738094086287847, + "raw_apical_rms": 0.160350078362088, + "teaching_rms": 0.0027380940860072885, + "traffic_rms": 0.16032627262381183 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.384102135825742e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.514405910573417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006609140139783976, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.32405499139785765 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.098513126373291, + "mean_feedback_forward_cosine": 0.825629183920947, + "mean_feedback_forward_relative_error": 0.5577839160507375, + "min_feedback_forward_cosine": 0.4139281213283539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002806733872328214, + "instruction_rms": 0.002806733871476678, + "raw_apical_rms": 0.16137592009991242, + "teaching_rms": 0.002806733872328214, + "traffic_rms": 0.16135106133359212 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.402244588978647e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5122284420077096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006833164944225116, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31489995611190796 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0807279348373413, + "mean_feedback_forward_cosine": 0.8359038759361613, + "mean_feedback_forward_relative_error": 0.5396850214763121, + "min_feedback_forward_cosine": 0.43376851081848145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027126405252002935, + "instruction_rms": 0.002712640525776598, + "raw_apical_rms": 0.1614713695608004, + "teaching_rms": 0.0027126405252002935, + "traffic_rms": 0.16144816076638513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.129415908162628e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5209754067163376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007243929023428913, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31143629782676696 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0643980503082275, + "mean_feedback_forward_cosine": 0.8452983178875663, + "mean_feedback_forward_relative_error": 0.5226983774792064, + "min_feedback_forward_cosine": 0.45132529735565186 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027025649776800734, + "instruction_rms": 0.0027025649788819084, + "raw_apical_rms": 0.1619282005105999, + "teaching_rms": 0.0027025649776800734, + "traffic_rms": 0.16190523076048755 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.616452452590238e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5134851767211678e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006847339694932586, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29772738583564756 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0457823276519775, + "mean_feedback_forward_cosine": 0.854599083553661, + "mean_feedback_forward_relative_error": 0.5055396803400733, + "min_feedback_forward_cosine": 0.4712110757827759 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002761482782204626, + "instruction_rms": 0.0027614827828063375, + "raw_apical_rms": 0.16258257613656435, + "teaching_rms": 0.002761482782204626, + "traffic_rms": 0.1625586360347466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.46648885549439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.510967737137601e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006350087813880074, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.29446000383377074 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.026662826538086, + "mean_feedback_forward_cosine": 0.8632767005400224, + "mean_feedback_forward_relative_error": 0.48891182243824005, + "min_feedback_forward_cosine": 0.4909406304359436 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027294029851671073, + "instruction_rms": 0.002729402984734353, + "raw_apical_rms": 0.16345852326418697, + "teaching_rms": 0.0027294029851671073, + "traffic_rms": 0.1634353017690766 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.029981136634888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5219791969399757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0065914767230672595, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.2873349623298645 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0072821378707886, + "mean_feedback_forward_cosine": 0.8715605811639265, + "mean_feedback_forward_relative_error": 0.472716176238927, + "min_feedback_forward_cosine": 0.5108346939086914 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027412687801244375, + "instruction_rms": 0.002741268779740909, + "raw_apical_rms": 0.1637920463690306, + "teaching_rms": 0.0027412687801244375, + "traffic_rms": 0.1637685970984676 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.260398777158116e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513659095315303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006347078384476118, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2780431655406952 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9849993586540222, + "mean_feedback_forward_cosine": 0.8796194163235751, + "mean_feedback_forward_relative_error": 0.4566669481721791, + "min_feedback_forward_cosine": 0.5322908163070679 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002772720722504387, + "instruction_rms": 0.002772720721653245, + "raw_apical_rms": 0.16517825215258577, + "teaching_rms": 0.002772720722504387, + "traffic_rms": 0.1651544968565768 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.4016616957142105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5124009446383405e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006232119474522236, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27643853506088256 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9637517333030701, + "mean_feedback_forward_cosine": 0.8872257872061295, + "mean_feedback_forward_relative_error": 0.44102646708488463, + "min_feedback_forward_cosine": 0.5526212453842163 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027165305410477045, + "instruction_rms": 0.0027165305406895435, + "raw_apical_rms": 0.16552759880769918, + "teaching_rms": 0.0027165305410477045, + "traffic_rms": 0.16550480070143003 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.4904397472910205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5104235840156505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006151879557324416, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2654162586069107 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.944539487361908, + "mean_feedback_forward_cosine": 0.8941527670080012, + "mean_feedback_forward_relative_error": 0.42625711262226107, + "min_feedback_forward_cosine": 0.5705175399780273 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002712585016804926, + "instruction_rms": 0.002712585016648845, + "raw_apical_rms": 0.16654005679792938, + "teaching_rms": 0.002712585016804926, + "traffic_rms": 0.166517485394477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.368676335772761e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.506944724660411e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006120622290934922, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.2637014389514923 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9225064516067505, + "mean_feedback_forward_cosine": 0.9008544921875, + "mean_feedback_forward_relative_error": 0.411741810224273, + "min_feedback_forward_cosine": 0.5907073020935059 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027195632493860525, + "instruction_rms": 0.0027195632494493604, + "raw_apical_rms": 0.16716096719402668, + "teaching_rms": 0.0027195632493860525, + "traffic_rms": 0.16713831787247851 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.625273130111964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5099903609261106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058345445216743674, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2554851693439484 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9018487334251404, + "mean_feedback_forward_cosine": 0.907134747505188, + "mean_feedback_forward_relative_error": 0.39767970808527686, + "min_feedback_forward_cosine": 0.6090466380119324 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026915925703477647, + "instruction_rms": 0.0026915925705541096, + "raw_apical_rms": 0.16747368530356588, + "teaching_rms": 0.0026915925703477647, + "traffic_rms": 0.167451493246592 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.8205895214769043e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.507205419149286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0059935289922810634, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25527560507774355 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8779973983764648, + "mean_feedback_forward_cosine": 0.9132803266698664, + "mean_feedback_forward_relative_error": 0.3836985009637746, + "min_feedback_forward_cosine": 0.629700779914856 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002709336798875379, + "instruction_rms": 0.002709336798922376, + "raw_apical_rms": 0.16876696760186286, + "teaching_rms": 0.002709336798875379, + "traffic_rms": 0.16874474387579133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.7212787396565545e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5155106164427826e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005770719968114469, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24738198353767396 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8550582528114319, + "mean_feedback_forward_cosine": 0.9191248839551752, + "mean_feedback_forward_relative_error": 0.3699219751087102, + "min_feedback_forward_cosine": 0.6488503813743591 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744658580348916, + "instruction_rms": 0.0027446585809002867, + "raw_apical_rms": 0.1692784749870594, + "teaching_rms": 0.002744658580348916, + "traffic_rms": 0.16925564762653938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 1.0283056184334782e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.503462783355452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005984480415672107, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24656046650886534 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8322657942771912, + "mean_feedback_forward_cosine": 0.924671167677099, + "mean_feedback_forward_relative_error": 0.35647109056060966, + "min_feedback_forward_cosine": 0.6674500107765198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002716551682468488, + "instruction_rms": 0.002716551681202091, + "raw_apical_rms": 0.16999963073479868, + "teaching_rms": 0.002716551682468488, + "traffic_rms": 0.16997738192670953 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.883178199108443e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.503999784704528e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005998146009132028, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.24410252239227295 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8098189830780029, + "mean_feedback_forward_cosine": 0.9298397335139188, + "mean_feedback_forward_relative_error": 0.343571529740637, + "min_feedback_forward_cosine": 0.6851625442504883 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002704993410636537, + "instruction_rms": 0.00270499340952124, + "raw_apical_rms": 0.17004736791511776, + "teaching_rms": 0.002704993410636537, + "traffic_rms": 0.17002524441399822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.865217810376635e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5008694971792143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00595464802904857, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.2346136785888672 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.788639485836029, + "mean_feedback_forward_cosine": 0.9345328157598322, + "mean_feedback_forward_relative_error": 0.3313400975682519, + "min_feedback_forward_cosine": 0.7013775706291199 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027144490760981355, + "instruction_rms": 0.0027144490765332522, + "raw_apical_rms": 0.1704820113064447, + "teaching_rms": 0.0027144490760981355, + "traffic_rms": 0.17045976833896143 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.369202054656853e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5075531941569833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005941322468501192, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22998918856143952 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7658517956733704, + "mean_feedback_forward_cosine": 0.9390178973024542, + "mean_feedback_forward_relative_error": 0.31936303905465385, + "min_feedback_forward_cosine": 0.7179839015007019 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027388282564563614, + "instruction_rms": 0.0027388282566992773, + "raw_apical_rms": 0.17149561445755873, + "teaching_rms": 0.0027388282564563614, + "traffic_rms": 0.1714732163783227 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.646818289686053e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5039550853157958e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058671058650391975, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.228340030670166 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7429981827735901, + "mean_feedback_forward_cosine": 0.9432894836772572, + "mean_feedback_forward_relative_error": 0.30768744457851754, + "min_feedback_forward_cosine": 0.7345595955848694 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027492090132646358, + "instruction_rms": 0.0027492090135381726, + "raw_apical_rms": 0.17173264824442055, + "teaching_rms": 0.0027492090132646358, + "traffic_rms": 0.17171000743235323 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.478892800032554e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4961528678032153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005707809539607657, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22297325903892518 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7205039262771606, + "mean_feedback_forward_cosine": 0.9472182219678705, + "mean_feedback_forward_relative_error": 0.29652146344835106, + "min_feedback_forward_cosine": 0.7503696084022522 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002699274124081228, + "instruction_rms": 0.0026992741242085087, + "raw_apical_rms": 0.17213260750946038, + "teaching_rms": 0.002699274124081228, + "traffic_rms": 0.17211085907920048 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 8.331967649155558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5001326548704666e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005755646294036971, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22228818331718445 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6984027028083801, + "mean_feedback_forward_cosine": 0.9509410045363687, + "mean_feedback_forward_relative_error": 0.2856178346005353, + "min_feedback_forward_cosine": 0.7656651735305786 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00271700873409705, + "instruction_rms": 0.002717008734128077, + "raw_apical_rms": 0.17280180485978924, + "teaching_rms": 0.00271700873409705, + "traffic_rms": 0.17277982802284342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1747259520689113e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4984288504050477e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005985720429172296, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22128331359386444 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6764211058616638, + "mean_feedback_forward_cosine": 0.9543936209245162, + "mean_feedback_forward_relative_error": 0.27514181313189595, + "min_feedback_forward_cosine": 0.7800532579421997 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002709935901320221, + "instruction_rms": 0.00270993590110685, + "raw_apical_rms": 0.17309379917053339, + "teaching_rms": 0.002709935901320221, + "traffic_rms": 0.173071956657856 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.971801980171222e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.499260260947927e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833380762116176, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21470431671142579 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6528832316398621, + "mean_feedback_forward_cosine": 0.9577399297194047, + "mean_feedback_forward_relative_error": 0.26469197462905536, + "min_feedback_forward_cosine": 0.7951393723487854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027628235506417013, + "instruction_rms": 0.0027628235505700546, + "raw_apical_rms": 0.1738290779543857, + "teaching_rms": 0.0027628235506417013, + "traffic_rms": 0.17380649136716134 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.574178084723826e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5033849938601963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005856943721158105, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2135639434337616 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6299036741256714, + "mean_feedback_forward_cosine": 0.9609023495153948, + "mean_feedback_forward_relative_error": 0.2544504498216239, + "min_feedback_forward_cosine": 0.8091050386428833 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027536471840989255, + "instruction_rms": 0.002753647183624104, + "raw_apical_rms": 0.174755363669475, + "teaching_rms": 0.0027536471840989255, + "traffic_rms": 0.17473309632877218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.033400540810362e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.497529781374043e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006092486563079558, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.21429219744205474 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6096960306167603, + "mean_feedback_forward_cosine": 0.9636846466497941, + "mean_feedback_forward_relative_error": 0.24503131820396942, + "min_feedback_forward_cosine": 0.8211265802383423 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027462287503163567, + "instruction_rms": 0.002746228750504722, + "raw_apical_rms": 0.17501511406704992, + "teaching_rms": 0.0027462287503163567, + "traffic_rms": 0.17499293449270764 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 4.459483093910421e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.496827060255142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006133579477802447, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20507440061092377 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.58868408203125, + "mean_feedback_forward_cosine": 0.9663694229992953, + "mean_feedback_forward_relative_error": 0.23567482084035873, + "min_feedback_forward_cosine": 0.8330232501029968 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002788761182192513, + "instruction_rms": 0.0027887611818201625, + "raw_apical_rms": 0.17594458221082196, + "teaching_rms": 0.002788761182192513, + "traffic_rms": 0.17592183768676548 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.44173860312003e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5080592249893096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006280252800208818, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.2034974954175949 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5699996948242188, + "mean_feedback_forward_cosine": 0.9687195626172153, + "mean_feedback_forward_relative_error": 0.22712430534037678, + "min_feedback_forward_cosine": 0.8433349132537842 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026731575028941345, + "instruction_rms": 0.002673157502333831, + "raw_apical_rms": 0.17700813020256884, + "teaching_rms": 0.0026731575028941345, + "traffic_rms": 0.1769873235663447 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.2035066033795374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5017710679382425e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00610103915669816, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.19972688953638076 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.550325870513916, + "mean_feedback_forward_cosine": 0.9709732543338429, + "mean_feedback_forward_relative_error": 0.21872749497944657, + "min_feedback_forward_cosine": 0.8538931608200073 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002717835118428842, + "instruction_rms": 0.0027178351175052664, + "raw_apical_rms": 0.17722203003458972, + "teaching_rms": 0.002717835118428842, + "traffic_rms": 0.1772006454723428 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 5.999399377287773e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4922741685155208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005937190934807443, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.1962828039741516 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5283382534980774, + "mean_feedback_forward_cosine": 0.9732055718248541, + "mean_feedback_forward_relative_error": 0.21019287739287723, + "min_feedback_forward_cosine": 0.8651987314224243 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002760480034646327, + "instruction_rms": 0.0027604800344301003, + "raw_apical_rms": 0.177716750738102, + "teaching_rms": 0.002760480034646327, + "traffic_rms": 0.17769463331364055 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 1.1262102361797588e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.499809594321995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006019799285633998, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.198746871881485 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5103389620780945, + "mean_feedback_forward_cosine": 0.9751089291139082, + "mean_feedback_forward_relative_error": 0.20251375328410756, + "min_feedback_forward_cosine": 0.8740372657775879 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002684252193379826, + "instruction_rms": 0.0026842521931177395, + "raw_apical_rms": 0.17827039607941997, + "teaching_rms": 0.002684252193379826, + "traffic_rms": 0.17824952594767854 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1688766932138606e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.498274480094141e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005988714848234756, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19467794558525087 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.49154627323150635, + "mean_feedback_forward_cosine": 0.9769123283299532, + "mean_feedback_forward_relative_error": 0.1950347569178451, + "min_feedback_forward_cosine": 0.8830726742744446 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002664375140555376, + "instruction_rms": 0.002664375140456436, + "raw_apical_rms": 0.17924584341280653, + "teaching_rms": 0.002664375140555376, + "traffic_rms": 0.17922537098901548 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.832449190279476e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4884402453295137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006116108332319452, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19025342669487 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4726116955280304, + "mean_feedback_forward_cosine": 0.9786086906086314, + "mean_feedback_forward_relative_error": 0.18781283477490598, + "min_feedback_forward_cosine": 0.8918552994728088 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00265934108324626, + "instruction_rms": 0.0026593410829282797, + "raw_apical_rms": 0.17989370183653353, + "teaching_rms": 0.00265934108324626, + "traffic_rms": 0.17987337427686068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.191707140918879e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4890153971951208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005906286966988396, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.18266307161092757 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4537387788295746, + "mean_feedback_forward_cosine": 0.9802491524002769, + "mean_feedback_forward_relative_error": 0.18047235039147463, + "min_feedback_forward_cosine": 0.9002604484558105 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027143535690182896, + "instruction_rms": 0.002714353568681626, + "raw_apical_rms": 0.17958034591684222, + "teaching_rms": 0.0027143535690182896, + "traffic_rms": 0.17955922788121503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.349976272053027e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.499637616653392e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005936952527896943, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.18880251668930054 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43645980954170227, + "mean_feedback_forward_cosine": 0.9817448681051081, + "mean_feedback_forward_relative_error": 0.17356848791241647, + "min_feedback_forward_cosine": 0.9076414108276367 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026550620897831515, + "instruction_rms": 0.002655062090175433, + "raw_apical_rms": 0.17980331039513187, + "teaching_rms": 0.0026550620897831515, + "traffic_rms": 0.17978306629196117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.297108422539168e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4932539674891137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005965728970932545, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.18843077926635743 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4198605418205261, + "mean_feedback_forward_cosine": 0.9831109122796492, + "mean_feedback_forward_relative_error": 0.16701087118549782, + "min_feedback_forward_cosine": 0.914421558380127 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026392736202412405, + "instruction_rms": 0.002639273620609916, + "raw_apical_rms": 0.18019670041627758, + "teaching_rms": 0.0026392736202412405, + "traffic_rms": 0.1801767993466816 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.338590452514836e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4979112706185428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005684992921142866, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.1808064446258545 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40299341082572937, + "mean_feedback_forward_cosine": 0.9844114715402776, + "mean_feedback_forward_relative_error": 0.16050030040470037, + "min_feedback_forward_cosine": 0.9210730791091919 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027142568524102014, + "instruction_rms": 0.002714256852345905, + "raw_apical_rms": 0.18032538911979093, + "teaching_rms": 0.0027142568524102014, + "traffic_rms": 0.18030426226248156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 8.407894114270675e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4923075430215646e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057010398688956115, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18632859456062317 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3866041302680969, + "mean_feedback_forward_cosine": 0.9855976007201455, + "mean_feedback_forward_relative_error": 0.15433649508790537, + "min_feedback_forward_cosine": 0.9272903203964233 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027006490642498765, + "instruction_rms": 0.002700649063897837, + "raw_apical_rms": 0.18080862530846392, + "teaching_rms": 0.0027006490642498765, + "traffic_rms": 0.18078785494432892 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.155787867281106e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5019170164792197e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0061135259567479654, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1800046843624115 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36960187554359436, + "mean_feedback_forward_cosine": 0.9867526943033392, + "mean_feedback_forward_relative_error": 0.14811216795986348, + "min_feedback_forward_cosine": 0.9334717392921448 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027252355608856633, + "instruction_rms": 0.002725235562080899, + "raw_apical_rms": 0.1812811856149048, + "teaching_rms": 0.0027252355608856633, + "traffic_rms": 0.18126004610690769 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.862533297180562e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4971890764273145e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005988503268362011, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18468582039356232 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35576748847961426, + "mean_feedback_forward_cosine": 0.9877345540306784, + "mean_feedback_forward_relative_error": 0.14254499599337578, + "min_feedback_forward_cosine": 0.9383275508880615 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026626480948363164, + "instruction_rms": 0.002662648093476927, + "raw_apical_rms": 0.18187451383364642, + "teaching_rms": 0.0026626480948363164, + "traffic_rms": 0.18185426421311832 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.054628005041195e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.491814491700462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0060516355778928495, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.17379911204338075 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.341165691614151, + "mean_feedback_forward_cosine": 0.9886939211325212, + "mean_feedback_forward_relative_error": 0.13691989874297922, + "min_feedback_forward_cosine": 0.9432326555252075 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002742297542302723, + "instruction_rms": 0.0027422975431780046, + "raw_apical_rms": 0.18213827013110045, + "teaching_rms": 0.002742297542302723, + "traffic_rms": 0.18211689684701102 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.150603499259887e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4984150646015722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005957774052336889, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.17903558188438415 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3280952274799347, + "mean_feedback_forward_cosine": 0.9895437955856323, + "mean_feedback_forward_relative_error": 0.13170490163293752, + "min_feedback_forward_cosine": 0.947439432144165 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026945714398406303, + "instruction_rms": 0.002694571439876451, + "raw_apical_rms": 0.18240486444450613, + "teaching_rms": 0.0026945714398406303, + "traffic_rms": 0.18238433186518632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.45010045982508e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5030875347773177e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005924772977991581, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17925691761016846 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3144298791885376, + "mean_feedback_forward_cosine": 0.9903578942472284, + "mean_feedback_forward_relative_error": 0.12659529240971262, + "min_feedback_forward_cosine": 0.9516677856445312 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002672899015007783, + "instruction_rms": 0.0026728990151781056, + "raw_apical_rms": 0.18344595300720584, + "teaching_rms": 0.002672899015007783, + "traffic_rms": 0.1834258679288141 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.787651329887012e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5011893892438375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006007994333805862, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.17378512043476105 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.300890177488327, + "mean_feedback_forward_cosine": 0.9911255684765903, + "mean_feedback_forward_relative_error": 0.1215421838177876, + "min_feedback_forward_cosine": 0.955715000629425 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002714685658448304, + "instruction_rms": 0.0027146856580466806, + "raw_apical_rms": 0.18382709214583234, + "teaching_rms": 0.002714685658448304, + "traffic_rms": 0.18380641156425492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.8219693954697505e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4926006548927225e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006097633915270201, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.17609858583450316 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2895379662513733, + "mean_feedback_forward_cosine": 0.9917840307409113, + "mean_feedback_forward_relative_error": 0.1169719752262939, + "min_feedback_forward_cosine": 0.958964467048645 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002655206526901862, + "instruction_rms": 0.0026552065271331876, + "raw_apical_rms": 0.1843576559142445, + "teaching_rms": 0.002655206526901862, + "traffic_rms": 0.18433789288204339 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.002170576697608e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4923624582222507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005926291225303201, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.17187404872894288 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27829137444496155, + "mean_feedback_forward_cosine": 0.9924183303659613, + "mean_feedback_forward_relative_error": 0.11238479854708368, + "min_feedback_forward_cosine": 0.9620615839958191 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002710543486908072, + "instruction_rms": 0.0027105434866209047, + "raw_apical_rms": 0.18438718904472642, + "teaching_rms": 0.002710543486908072, + "traffic_rms": 0.1843666451236843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.896579035104878e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484495620178975e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005738775496578574, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17466742164611818 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2671208381652832, + "mean_feedback_forward_cosine": 0.992997576973655, + "mean_feedback_forward_relative_error": 0.10806869698519056, + "min_feedback_forward_cosine": 0.9650495052337646 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002660826974935695, + "instruction_rms": 0.0026608269755329836, + "raw_apical_rms": 0.1850226516700624, + "teaching_rms": 0.002660826974935695, + "traffic_rms": 0.18500279320360521 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.553655396146496e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4911308755548197e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005460609321289271, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.16644893995761872 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2560533285140991, + "mean_feedback_forward_cosine": 0.9935503992167386, + "mean_feedback_forward_relative_error": 0.10377158830789003, + "min_feedback_forward_cosine": 0.9678542613983154 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027544022459381288, + "instruction_rms": 0.0027544022461819958, + "raw_apical_rms": 0.18549669961122786, + "teaching_rms": 0.0027544022459381288, + "traffic_rms": 0.18547550483511785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.82698421630812e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4945048682973907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005745918348240271, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.17110243844985962 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24560950696468353, + "mean_feedback_forward_cosine": 0.9940380963412199, + "mean_feedback_forward_relative_error": 0.0998407182707028, + "min_feedback_forward_cosine": 0.9703977704048157 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00265093646446746, + "instruction_rms": 0.0026509364642112827, + "raw_apical_rms": 0.18568833720966033, + "teaching_rms": 0.00265093646446746, + "traffic_rms": 0.18566876172137756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.658588347363548e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484481315388553e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005600620307193533, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.16274750321388246 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23551318049430847, + "mean_feedback_forward_cosine": 0.9945091789419, + "mean_feedback_forward_relative_error": 0.09583028155294332, + "min_feedback_forward_cosine": 0.9727450609207153 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027441751521812695, + "instruction_rms": 0.002744175152484084, + "raw_apical_rms": 0.18679847675580416, + "teaching_rms": 0.0027441751521812695, + "traffic_rms": 0.1867776277802787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.766995852558466e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4839523385048985e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058507432804929885, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1717451312160492 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2254258245229721, + "mean_feedback_forward_cosine": 0.9949419931931929, + "mean_feedback_forward_relative_error": 0.09205600087615577, + "min_feedback_forward_cosine": 0.9750077724456787 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027330063941230285, + "instruction_rms": 0.0027330063942390837, + "raw_apical_rms": 0.18724200154112863, + "teaching_rms": 0.0027330063941230285, + "traffic_rms": 0.18722145144791583 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.0082177226422573e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.485527360168689e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005474347340255327, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16850471174240111 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21620775759220123, + "mean_feedback_forward_cosine": 0.9953409563411366, + "mean_feedback_forward_relative_error": 0.08840271156619896, + "min_feedback_forward_cosine": 0.9769967794418335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027384400922097715, + "instruction_rms": 0.0027384400932530767, + "raw_apical_rms": 0.1874413875391775, + "teaching_rms": 0.0027384400922097715, + "traffic_rms": 0.18742066599396165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.82139011334948e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4866016125743815e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057920571394057705, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.1602698998260498 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20743760466575623, + "mean_feedback_forward_cosine": 0.9956982590935447, + "mean_feedback_forward_relative_error": 0.08500193560665303, + "min_feedback_forward_cosine": 0.9788150787353516 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026415409205600366, + "instruction_rms": 0.0026415409203534765, + "raw_apical_rms": 0.1883252030729858, + "teaching_rms": 0.0026415409205600366, + "traffic_rms": 0.18830597472486543 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.010427083287141e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4891452249653445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005805818102755636, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.16418667131900788 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19822026789188385, + "mean_feedback_forward_cosine": 0.9960439064285972, + "mean_feedback_forward_relative_error": 0.08159297315234487, + "min_feedback_forward_cosine": 0.9806467294692993 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002670809836066248, + "instruction_rms": 0.002670809835063417, + "raw_apical_rms": 0.18818173302187235, + "teaching_rms": 0.002670809836066248, + "traffic_rms": 0.18816201965489845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.136671425518216e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.500931509103375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005907966424484982, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1620645476436615 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.190128892660141, + "mean_feedback_forward_cosine": 0.9963437058708885, + "mean_feedback_forward_relative_error": 0.07847126156769016, + "min_feedback_forward_cosine": 0.9821842908859253 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026653986992704933, + "instruction_rms": 0.002665398699077468, + "raw_apical_rms": 0.18917828664579017, + "teaching_rms": 0.0026653986992704933, + "traffic_rms": 0.18915890402351607 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.754499537464653e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480230018878317e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005596040555727568, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15636374557495117 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1819751113653183, + "mean_feedback_forward_cosine": 0.9966294906356118, + "mean_feedback_forward_relative_error": 0.0753943392160264, + "min_feedback_forward_cosine": 0.9836736917495728 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026968256156761016, + "instruction_rms": 0.0026968256153775327, + "raw_apical_rms": 0.18935510363067287, + "teaching_rms": 0.0026968256156761016, + "traffic_rms": 0.1893352194814942 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.918212301978201e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4768652011833755e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0055450689618138415, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15599127224445344 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17416630685329437, + "mean_feedback_forward_cosine": 0.9968989545648749, + "mean_feedback_forward_relative_error": 0.07235846170647578, + "min_feedback_forward_cosine": 0.9850413203239441 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027180539313595643, + "instruction_rms": 0.0027180539316359964, + "raw_apical_rms": 0.18968632248078213, + "teaching_rms": 0.0027180539313595643, + "traffic_rms": 0.189666169677182 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.398345047704424e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4802031950066428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005681046297481906, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16048648918867112 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16696111857891083, + "mean_feedback_forward_cosine": 0.997138261795044, + "mean_feedback_forward_relative_error": 0.06955248916691, + "min_feedback_forward_cosine": 0.986241340637207 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026404591563534375, + "instruction_rms": 0.0026404591568132255, + "raw_apical_rms": 0.19031131551390063, + "teaching_rms": 0.0026404591563534375, + "traffic_rms": 0.1902923191833995 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.384455275565415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4858139683994578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005328538068270849, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15991598387241363 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16049210727214813, + "mean_feedback_forward_cosine": 0.9973547902974216, + "mean_feedback_forward_relative_error": 0.06688470335846597, + "min_feedback_forward_cosine": 0.9872801303863525 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002639214064001048, + "instruction_rms": 0.0026392140638360397, + "raw_apical_rms": 0.19106928072143337, + "teaching_rms": 0.002639214064001048, + "traffic_rms": 0.19105039286851047 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.0731622397725e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4715494178358734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00556611939160749, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.15247124420642852 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15385060012340546, + "mean_feedback_forward_cosine": 0.997557439587333, + "mean_feedback_forward_relative_error": 0.06429105201228098, + "min_feedback_forward_cosine": 0.9883031845092773 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026681067989476346, + "instruction_rms": 0.0026681067998364726, + "raw_apical_rms": 0.19117395586662655, + "teaching_rms": 0.0026681067989476346, + "traffic_rms": 0.1911546044770046 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.939963270724547e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4827310844934857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0055671561122739004, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15150256306648255 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14684991538524628, + "mean_feedback_forward_cosine": 0.997753357887268, + "mean_feedback_forward_relative_error": 0.06170663378116759, + "min_feedback_forward_cosine": 0.9893401265144348 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002720927986619134, + "instruction_rms": 0.002720927987159682, + "raw_apical_rms": 0.19116774390199975, + "teaching_rms": 0.002720927986619134, + "traffic_rms": 0.1911476161593291 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.76103582009091e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.48429502964912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005505891978268893, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15532715906620026 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14046870172023773, + "mean_feedback_forward_cosine": 0.9979327949610624, + "mean_feedback_forward_relative_error": 0.059233984401957554, + "min_feedback_forward_cosine": 0.9902418255805969 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027356576089698115, + "instruction_rms": 0.0027356576084319527, + "raw_apical_rms": 0.19162561696523955, + "teaching_rms": 0.0027356576089698115, + "traffic_rms": 0.19160540444431545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.035745935330851e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4950216628707212e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005554710419520541, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.15724634010314942 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1347571462392807, + "mean_feedback_forward_cosine": 0.9980938857251948, + "mean_feedback_forward_relative_error": 0.05692117818715897, + "min_feedback_forward_cosine": 0.991012454032898 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027125745833637806, + "instruction_rms": 0.0027125745829940104, + "raw_apical_rms": 0.1917939404331772, + "teaching_rms": 0.0027125745833637806, + "traffic_rms": 0.1917740754563628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.476822278500549e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.48663609317214e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005388191210017593, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15491003794789315 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12929846346378326, + "mean_feedback_forward_cosine": 0.9982388453050093, + "mean_feedback_forward_relative_error": 0.0547489713200114, + "min_feedback_forward_cosine": 0.9917217493057251 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026528897508248253, + "instruction_rms": 0.0026528897515970874, + "raw_apical_rms": 0.1926668455164547, + "teaching_rms": 0.0026528897508248253, + "traffic_rms": 0.19264781811201465 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 8.286859554118564e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4829239017387906e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00553722700393831, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1501428285551071 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12421248108148575, + "mean_feedback_forward_cosine": 0.9983759078112515, + "mean_feedback_forward_relative_error": 0.05260178695686839, + "min_feedback_forward_cosine": 0.9923573136329651 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002697842206245549, + "instruction_rms": 0.0026978422064576875, + "raw_apical_rms": 0.193514785769868, + "teaching_rms": 0.002697842206245549, + "traffic_rms": 0.19349529821164013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.172270161641791e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480632409701656e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005483752235576802, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15325217506408692 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11903703957796097, + "mean_feedback_forward_cosine": 0.9984997803514654, + "mean_feedback_forward_relative_error": 0.050592693652619015, + "min_feedback_forward_cosine": 0.9929812550544739 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002685146277560296, + "instruction_rms": 0.00268514627669799, + "raw_apical_rms": 0.19386935073996772, + "teaching_rms": 0.002685146277560296, + "traffic_rms": 0.19384998785456464 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.306372313247266e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.483001605875091e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0055556696070504655, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14831138321876525 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11387547850608826, + "mean_feedback_forward_cosine": 0.9986178744922984, + "mean_feedback_forward_relative_error": 0.0485988156192682, + "min_feedback_forward_cosine": 0.9935735464096069 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026892984516622065, + "instruction_rms": 0.002689298452234876, + "raw_apical_rms": 0.19381157601006951, + "teaching_rms": 0.0026892984516622065, + "traffic_rms": 0.19379222099141588 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.864793450882644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4805781571390814e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005673575025965466, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1475057663822174 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10914173722267151, + "mean_feedback_forward_cosine": 0.9987257047133012, + "mean_feedback_forward_relative_error": 0.046692649495195256, + "min_feedback_forward_cosine": 0.9940951466560364 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002707446801658122, + "instruction_rms": 0.0027074468008335953, + "raw_apical_rms": 0.19490610957578813, + "teaching_rms": 0.002707446801658122, + "traffic_rms": 0.19488656025149476 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.8793279578139845e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4706892809021093e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005191779896525374, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.14866874879837036 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1046086773276329, + "mean_feedback_forward_cosine": 0.9988254601305181, + "mean_feedback_forward_relative_error": 0.04485084143551913, + "min_feedback_forward_cosine": 0.9945739507675171 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027242884675772253, + "instruction_rms": 0.0027242884675424206, + "raw_apical_rms": 0.19540975368732447, + "teaching_rms": 0.0027242884675772253, + "traffic_rms": 0.19538998278133352 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.180718669502312e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474773803523709e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005246561816305377, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14888891705036164 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10021883249282837, + "mean_feedback_forward_cosine": 0.9989158489487387, + "mean_feedback_forward_relative_error": 0.04311930025843057, + "min_feedback_forward_cosine": 0.9950194358825684 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002693345890631095, + "instruction_rms": 0.0026933458901430467, + "raw_apical_rms": 0.19556882416054872, + "teaching_rms": 0.002693345890631095, + "traffic_rms": 0.1955495263338322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.602576974046315e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.470937748647442e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005078910160794891, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.14472049828052522 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09581611305475235, + "mean_feedback_forward_cosine": 0.9990024891766635, + "mean_feedback_forward_relative_error": 0.04139081283726476, + "min_feedback_forward_cosine": 0.995445728302002 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002742003564110486, + "instruction_rms": 0.002742003564704616, + "raw_apical_rms": 0.19603411154988898, + "teaching_rms": 0.002742003564110486, + "traffic_rms": 0.1960141425878438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.071874819075674e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.468354803251202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00511159262017984, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.15269218800067902 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0917789563536644, + "mean_feedback_forward_cosine": 0.9990798885172064, + "mean_feedback_forward_relative_error": 0.03978007679635828, + "min_feedback_forward_cosine": 0.9958189725875854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002689767271455075, + "instruction_rms": 0.002689767271721301, + "raw_apical_rms": 0.19673704843669637, + "teaching_rms": 0.002689767271455075, + "traffic_rms": 0.19671792379262165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.931858382268729e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4791794750082213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005167129442606366, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14823474292993546 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08791062235832214, + "mean_feedback_forward_cosine": 0.9991524934768676, + "mean_feedback_forward_relative_error": 0.03820779186419465, + "min_feedback_forward_cosine": 0.9961642026901245 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00270110678096097, + "instruction_rms": 0.002701106781420777, + "raw_apical_rms": 0.19736663528167958, + "teaching_rms": 0.00270110678096097, + "traffic_rms": 0.1973474663415084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.916197665101208e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4763722753244483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00532267559677848, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.14822878425598143 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0846085399389267, + "mean_feedback_forward_cosine": 0.9992151975631713, + "mean_feedback_forward_relative_error": 0.036782165498218757, + "min_feedback_forward_cosine": 0.9964458346366882 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026655710020981327, + "instruction_rms": 0.0026655710014431384, + "raw_apical_rms": 0.19832052232170086, + "teaching_rms": 0.0026655710020981327, + "traffic_rms": 0.1983018833502957 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.293979275071251e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4721804704192416e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005423458451203647, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14442641533851625 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08136910945177078, + "mean_feedback_forward_cosine": 0.9992750623009422, + "mean_feedback_forward_relative_error": 0.03535802076486024, + "min_feedback_forward_cosine": 0.9967116713523865 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00272810964782141, + "instruction_rms": 0.0027281096480237456, + "raw_apical_rms": 0.198665116699432, + "teaching_rms": 0.00272810964782141, + "traffic_rms": 0.19864570828223777 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 1.0495469382476585e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4756904906268807e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005524084704260759, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1430003475189209 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0777994841337204, + "mean_feedback_forward_cosine": 0.9993326013738458, + "mean_feedback_forward_relative_error": 0.03394931190731851, + "min_feedback_forward_cosine": 0.996992290019989 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027240797588575566, + "instruction_rms": 0.002724079758267756, + "raw_apical_rms": 0.19863590530231257, + "teaching_rms": 0.0027240797588575566, + "traffic_rms": 0.1986164375353274 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.082190478513439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4694126723778814e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005264208722605274, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.14785423511981965 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07459433376789093, + "mean_feedback_forward_cosine": 0.9993848963217302, + "mean_feedback_forward_relative_error": 0.032613970136100595, + "min_feedback_forward_cosine": 0.9972344636917114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026739769661550086, + "instruction_rms": 0.0026739769670374816, + "raw_apical_rms": 0.1990884431705579, + "teaching_rms": 0.0026739769661550086, + "traffic_rms": 0.19906986230559304 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.161327803511085e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474640909338308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00532078878628481, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.14291324827194213 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07143103331327438, + "mean_feedback_forward_cosine": 0.9994342056187716, + "mean_feedback_forward_relative_error": 0.03129539456726475, + "min_feedback_forward_cosine": 0.9974645376205444 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027373663419869598, + "instruction_rms": 0.0027373663419299104, + "raw_apical_rms": 0.19921440341597585, + "teaching_rms": 0.0027373663419869598, + "traffic_rms": 0.1991947927604281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 1.023156102064604e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4671252830762428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005176140216877374, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14808677036762238 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07140765339136124, + "mean_feedback_forward_cosine": 0.9994345762512901, + "mean_feedback_forward_relative_error": 0.031284770123999224, + "min_feedback_forward_cosine": 0.9974661469459534 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018495065379233345, + "instruction_rms": 0.0018495065373225433, + "raw_apical_rms": 0.19942685833346657, + "teaching_rms": 0.0018495065379233345, + "traffic_rms": 0.19941762747600783 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.28745987998991e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.458424118525584e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.004929282239103567, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07751537271499634 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07139676809310913, + "mean_feedback_forward_cosine": 0.9994347463954579, + "mean_feedback_forward_relative_error": 0.031279056587002495, + "min_feedback_forward_cosine": 0.9974669814109802 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015777643795577298, + "instruction_rms": 0.0015777643791070927, + "raw_apical_rms": 0.19941121707114334, + "teaching_rms": 0.0015777643795577298, + "traffic_rms": 0.19940438493955723 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.394533696469807e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4592925348359475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005013067033939708, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.050618355315923694 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07138567417860031, + "mean_feedback_forward_cosine": 0.9994349122047425, + "mean_feedback_forward_relative_error": 0.0312735039148141, + "min_feedback_forward_cosine": 0.9974678158760071 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001565832305922256, + "instruction_rms": 0.001565832305939745, + "raw_apical_rms": 0.19934545364122225, + "teaching_rms": 0.001565832305922256, + "traffic_rms": 0.19933869213206315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.948245356852567e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4631476323350904e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005049516117945999, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04441148671329021 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07137410342693329, + "mean_feedback_forward_cosine": 0.9994351300326261, + "mean_feedback_forward_relative_error": 0.031267848864875054, + "min_feedback_forward_cosine": 0.9974687099456787 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015625445229474767, + "instruction_rms": 0.0015625445225821968, + "raw_apical_rms": 0.1993275369581802, + "teaching_rms": 0.0015625445229474767, + "traffic_rms": 0.19932066886211822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.269783873833148e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4633837143850747e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005009128719219743, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03961767840951681 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07136402279138565, + "mean_feedback_forward_cosine": 0.999435276334936, + "mean_feedback_forward_relative_error": 0.03126273304224014, + "min_feedback_forward_cosine": 0.9974693059921265 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014598623118464557, + "instruction_rms": 0.00145986231204941, + "raw_apical_rms": 0.1992605119956912, + "teaching_rms": 0.0014598623118464557, + "traffic_rms": 0.1992543425729492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.56562386072801e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.457785573289802e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00498740630431355, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03284155424594879 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07135354727506638, + "mean_feedback_forward_cosine": 0.9994354551488703, + "mean_feedback_forward_relative_error": 0.03125756416808475, + "min_feedback_forward_cosine": 0.9974699020385742 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014555804135443797, + "instruction_rms": 0.0014555804137796576, + "raw_apical_rms": 0.1992600566467235, + "teaching_rms": 0.0014555804135443797, + "traffic_rms": 0.19925389700715707 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.907467309247295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4549661306616943e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051395152897739434, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.03039330755710602 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0713423490524292, + "mean_feedback_forward_cosine": 0.9994356404651296, + "mean_feedback_forward_relative_error": 0.031252134692939845, + "min_feedback_forward_cosine": 0.9974709153175354 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014833573690659213, + "instruction_rms": 0.001483357369263451, + "raw_apical_rms": 0.19920526037322447, + "teaching_rms": 0.0014833573690659213, + "traffic_rms": 0.1991988299909125 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.633653898939883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4574680997490702e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005082147892737552, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.02948811733663082 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07133160531520844, + "mean_feedback_forward_cosine": 0.9994358344511552, + "mean_feedback_forward_relative_error": 0.03124694436449896, + "min_feedback_forward_cosine": 0.9974715709686279 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001438791539101049, + "instruction_rms": 0.001438791539086446, + "raw_apical_rms": 0.19910229308583682, + "teaching_rms": 0.001438791539101049, + "traffic_rms": 0.1990960171708098 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.517689854490527e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4567295500488302e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005088590825404236, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.026679220786094665 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07132166624069214, + "mean_feedback_forward_cosine": 0.999435999176719, + "mean_feedback_forward_relative_error": 0.031241951772773807, + "min_feedback_forward_cosine": 0.9974722862243652 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014028609862289602, + "instruction_rms": 0.0014028609863941414, + "raw_apical_rms": 0.19911387872688596, + "teaching_rms": 0.0014028609862289602, + "traffic_rms": 0.19910790504564585 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.577410260888428e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.456873033808295e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005129392407500083, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.02497885542213917 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07131074368953705, + "mean_feedback_forward_cosine": 0.9994361909953031, + "mean_feedback_forward_relative_error": 0.031236791907047683, + "min_feedback_forward_cosine": 0.9974730014801025 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001355116720174963, + "instruction_rms": 0.0013551167201685502, + "raw_apical_rms": 0.19905342671168053, + "teaching_rms": 0.001355116720174963, + "traffic_rms": 0.19904777532813817 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.8555411186489437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.461635199129888e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005135905031312674, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.022831981168985365 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07130035758018494, + "mean_feedback_forward_cosine": 0.999436354637146, + "mean_feedback_forward_relative_error": 0.031231847007504917, + "min_feedback_forward_cosine": 0.9974737763404846 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013172750117069164, + "instruction_rms": 0.0013172750116959972, + "raw_apical_rms": 0.1990287998584814, + "teaching_rms": 0.0013172750117069164, + "traffic_rms": 0.19902338053783347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.909274219424333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4635918359050652e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005082521198541561, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02103618494540453 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0712905153632164, + "mean_feedback_forward_cosine": 0.9994365345348012, + "mean_feedback_forward_relative_error": 0.031226877952841195, + "min_feedback_forward_cosine": 0.9974745512008667 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001322208176314824, + "instruction_rms": 0.0013222081758442656, + "raw_apical_rms": 0.19900189212558503, + "teaching_rms": 0.001322208176314824, + "traffic_rms": 0.19899650148956916 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.730419244149743e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4621691387073312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005223045990181792, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01949120617210865 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07128015905618668, + "mean_feedback_forward_cosine": 0.9994367339394309, + "mean_feedback_forward_relative_error": 0.031221711787987837, + "min_feedback_forward_cosine": 0.9974753260612488 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013203694610326472, + "instruction_rms": 0.0013203694614501908, + "raw_apical_rms": 0.19900873173707467, + "teaching_rms": 0.0013203694610326472, + "traffic_rms": 0.19900313380932755 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.530133754902636e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4583809279651156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005033622846982757, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.019819492907971144 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07126913219690323, + "mean_feedback_forward_cosine": 0.9994369170882486, + "mean_feedback_forward_relative_error": 0.031216473183171316, + "min_feedback_forward_cosine": 0.9974759817123413 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013172640447274018, + "instruction_rms": 0.0013172640443568607, + "raw_apical_rms": 0.1989699759219127, + "teaching_rms": 0.0013172640447274018, + "traffic_rms": 0.19896426031619208 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 8.568976284095697e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4627477679988265e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005037033788794907, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.018879418965876103 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07125793397426605, + "mean_feedback_forward_cosine": 0.9994371154091575, + "mean_feedback_forward_relative_error": 0.031211044707081536, + "min_feedback_forward_cosine": 0.9974768757820129 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001313282273760199, + "instruction_rms": 0.0013132822738414164, + "raw_apical_rms": 0.19889969451394604, + "teaching_rms": 0.001313282273760199, + "traffic_rms": 0.19889414060284477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.973375039047824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.460488534074445e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005099359185110131, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.01721682056903839 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07124721258878708, + "mean_feedback_forward_cosine": 0.9994373072277416, + "mean_feedback_forward_relative_error": 0.031205538186160002, + "min_feedback_forward_cosine": 0.997477650642395 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013163457189959336, + "instruction_rms": 0.0013163457189392684, + "raw_apical_rms": 0.19881240929937205, + "teaching_rms": 0.0013163457189959336, + "traffic_rms": 0.1988066195071072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.152397785148423e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4593228303730974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005070992993632505, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.016983157099336385 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07123630493879318, + "mean_feedback_forward_cosine": 0.9994375131346963, + "mean_feedback_forward_relative_error": 0.031200243498791348, + "min_feedback_forward_cosine": 0.9974783658981323 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012491100280182322, + "instruction_rms": 0.0012491100283490205, + "raw_apical_rms": 0.19884305199908187, + "teaching_rms": 0.0012491100280182322, + "traffic_rms": 0.19883770739003226 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.351935155635147e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4569821454584868e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005110478302105733, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01545833678394556 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07122577726840973, + "mean_feedback_forward_cosine": 0.999437712539326, + "mean_feedback_forward_relative_error": 0.03119498146697879, + "min_feedback_forward_cosine": 0.9974790811538696 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012470871607667604, + "instruction_rms": 0.0012470871606689025, + "raw_apical_rms": 0.19881528212057398, + "teaching_rms": 0.0012470871607667604, + "traffic_rms": 0.1988100765851186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.465253221676903e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.460795219568578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005162518784697405, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.015599264223873616 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07121507078409195, + "mean_feedback_forward_cosine": 0.9994379032741894, + "mean_feedback_forward_relative_error": 0.031189677711914887, + "min_feedback_forward_cosine": 0.9974797964096069 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012191826979796856, + "instruction_rms": 0.001219182697865101, + "raw_apical_rms": 0.19879420848450352, + "teaching_rms": 0.0012191826979796856, + "traffic_rms": 0.1987890617499381 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.496817909848687e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4590689263176857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005080241773836865, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.014447662914395332 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07120535522699356, + "mean_feedback_forward_cosine": 0.9994380647485906, + "mean_feedback_forward_relative_error": 0.031184876092116943, + "min_feedback_forward_cosine": 0.9974804520606995 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011559709399052696, + "instruction_rms": 0.0011559709397655112, + "raw_apical_rms": 0.19879655592713513, + "teaching_rms": 0.0011559709399052696, + "traffic_rms": 0.19879184472646552 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.120876060502269e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.458782286472728e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005169515235283719, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.013233123388588428 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0711955726146698, + "mean_feedback_forward_cosine": 0.9994382554834539, + "mean_feedback_forward_relative_error": 0.03117991037328135, + "min_feedback_forward_cosine": 0.9974811673164368 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011546298416442813, + "instruction_rms": 0.0011546298417712396, + "raw_apical_rms": 0.19881600040907224, + "teaching_rms": 0.0011546298416442813, + "traffic_rms": 0.19881140316187348 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.087182834515261e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.459817538740061e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005079960235353323, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.01236043606929481 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0711865946650505, + "mean_feedback_forward_cosine": 0.9994384397159923, + "mean_feedback_forward_relative_error": 0.0311754064719108, + "min_feedback_forward_cosine": 0.9974818825721741 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010654257456607832, + "instruction_rms": 0.0010654257456061908, + "raw_apical_rms": 0.1987333250366317, + "teaching_rms": 0.0010654257456607832, + "traffic_rms": 0.19872916824189388 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.445023359447077e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4565052243745985e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051418137236353585, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.011380807536076754 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07117650657892227, + "mean_feedback_forward_cosine": 0.9994386271996931, + "mean_feedback_forward_relative_error": 0.03117037285965952, + "min_feedback_forward_cosine": 0.9974825382232666 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001100886365428987, + "instruction_rms": 0.0011008863646458282, + "raw_apical_rms": 0.1987470751676122, + "teaching_rms": 0.001100886365428987, + "traffic_rms": 0.19874242371306938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.404133839372662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4620149112774336e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005114624300211534, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.012123570803254843 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07116719335317612, + "mean_feedback_forward_cosine": 0.9994387875903736, + "mean_feedback_forward_relative_error": 0.031165640733458778, + "min_feedback_forward_cosine": 0.9974830150604248 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010821219510699267, + "instruction_rms": 0.0010821219514403917, + "raw_apical_rms": 0.19875239092199495, + "teaching_rms": 0.0010821219510699267, + "traffic_rms": 0.1987480129076099 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.525664986156698e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4564971951480037e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005143106475378592, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.011324955357462167 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07115576416254044, + "mean_feedback_forward_cosine": 0.9994390032508157, + "mean_feedback_forward_relative_error": 0.031160284036939795, + "min_feedback_forward_cosine": 0.9974840879440308 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011076464696770291, + "instruction_rms": 0.001107646469520249, + "raw_apical_rms": 0.1987321062949196, + "teaching_rms": 0.0011076464696770291, + "traffic_rms": 0.1987274947959835 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.856924272824472e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4564302447873456e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005090543421855355, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.011298456238918007 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07114609330892563, + "mean_feedback_forward_cosine": 0.9994391885670748, + "mean_feedback_forward_relative_error": 0.0311554063280875, + "min_feedback_forward_cosine": 0.9974846839904785 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010425381095331763, + "instruction_rms": 0.001042538109407465, + "raw_apical_rms": 0.1986928829977562, + "teaching_rms": 0.0010425381095331763, + "traffic_rms": 0.1986887193875803 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.911442795853873e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.457654979571983e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005189060147206698, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.010381140566021204 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07113548368215561, + "mean_feedback_forward_cosine": 0.9994393944740295, + "mean_feedback_forward_relative_error": 0.031150113994424994, + "min_feedback_forward_cosine": 0.9974853992462158 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010989234910148077, + "instruction_rms": 0.0010989234909653993, + "raw_apical_rms": 0.19866438239827142, + "teaching_rms": 0.0010989234910148077, + "traffic_rms": 0.19865962507319254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.175053357537763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4589869995995475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005153233335700686, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.010389255159944296 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07112447172403336, + "mean_feedback_forward_cosine": 0.9994395657019182, + "mean_feedback_forward_relative_error": 0.031144795151935384, + "min_feedback_forward_cosine": 0.9974862337112427 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010843093573535106, + "instruction_rms": 0.0010843093572660036, + "raw_apical_rms": 0.19862367092591732, + "teaching_rms": 0.0010843093573535106, + "traffic_rms": 0.19861922087980705 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.024796389032417e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4546360509790742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005191692607743386, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.010292481841444968 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07111339271068573, + "mean_feedback_forward_cosine": 0.9994397911158475, + "mean_feedback_forward_relative_error": 0.03113898594271053, + "min_feedback_forward_cosine": 0.9974867105484009 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001142533804928452, + "instruction_rms": 0.0011425338045973905, + "raw_apical_rms": 0.19856723720836575, + "teaching_rms": 0.001142533804928452, + "traffic_rms": 0.1985623881440384 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.337993679357169e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4637993288262236e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005282985700801881, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.0108459441678226 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07110411673784256, + "mean_feedback_forward_cosine": 0.9994399677623402, + "mean_feedback_forward_relative_error": 0.031134315808726984, + "min_feedback_forward_cosine": 0.9974874258041382 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009974149181451748, + "instruction_rms": 0.0009974149186470814, + "raw_apical_rms": 0.19850268094715104, + "teaching_rms": 0.0009974149181451748, + "traffic_rms": 0.19849871831388485 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.762687831951553e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4528464100260063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005229958752258195, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009308304391801357 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07109230011701584, + "mean_feedback_forward_cosine": 0.9994402083483609, + "mean_feedback_forward_relative_error": 0.03112850643863732, + "min_feedback_forward_cosine": 0.997488260269165 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011153761913156438, + "instruction_rms": 0.001115376191328265, + "raw_apical_rms": 0.19861464569274487, + "teaching_rms": 0.0011153761913156438, + "traffic_rms": 0.19860973834457943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1028347302044494e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4559822076352724e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005200775910773839, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.010205369082093239 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07108078896999359, + "mean_feedback_forward_cosine": 0.999440424008803, + "mean_feedback_forward_relative_error": 0.031122633603147484, + "min_feedback_forward_cosine": 0.9974890947341919 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010602016613338163, + "instruction_rms": 0.0010602016612856612, + "raw_apical_rms": 0.1986677366997766, + "teaching_rms": 0.0010602016613338163, + "traffic_rms": 0.19866331019454708 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.215456486799667e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.454069803832435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005192546472282864, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.009919193947017194 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07106922566890717, + "mean_feedback_forward_cosine": 0.9994406266645952, + "mean_feedback_forward_relative_error": 0.0311172089624134, + "min_feedback_forward_cosine": 0.9974900484085083 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009962794817005125, + "instruction_rms": 0.0009962794818440552, + "raw_apical_rms": 0.1985586627136116, + "teaching_rms": 0.0009962794817005125, + "traffic_rms": 0.19855464919534335 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.408561231159183e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4539696995937464e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005134946019578701, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.008672905013412236 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0710592120885849, + "mean_feedback_forward_cosine": 0.9994408206506209, + "mean_feedback_forward_relative_error": 0.0311122240498662, + "min_feedback_forward_cosine": 0.9974907040596008 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009583823850593243, + "instruction_rms": 0.0009583823853470254, + "raw_apical_rms": 0.19856861505706658, + "teaching_rms": 0.0009583823850593243, + "traffic_rms": 0.19856486841591087 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.916759159343201e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4541923442166532e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005115200526328582, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.008275501573011278 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0710485577583313, + "mean_feedback_forward_cosine": 0.9994410233064132, + "mean_feedback_forward_relative_error": 0.03110693013295531, + "min_feedback_forward_cosine": 0.9974914193153381 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009721696787663096, + "instruction_rms": 0.000972169678946262, + "raw_apical_rms": 0.1986525121878335, + "teaching_rms": 0.0009721696787663096, + "traffic_rms": 0.19864853084562625 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.8668117475235704e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4622831874983543e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005181926677309708, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.00862616381302476 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07103725522756577, + "mean_feedback_forward_cosine": 0.9994412237947637, + "mean_feedback_forward_relative_error": 0.031101475926962767, + "min_feedback_forward_cosine": 0.9974921941757202 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009819280440553045, + "instruction_rms": 0.0009819280445649934, + "raw_apical_rms": 0.19865248082503967, + "teaching_rms": 0.0009819280440553045, + "traffic_rms": 0.19864857092917637 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.376549149284983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.455719640269268e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005153495492990665, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008505770028084517 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0710277184844017, + "mean_feedback_forward_cosine": 0.9994414047761397, + "mean_feedback_forward_relative_error": 0.031096697818826544, + "min_feedback_forward_cosine": 0.9974929690361023 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009007894767279102, + "instruction_rms": 0.0009007894766342797, + "raw_apical_rms": 0.1986070899519479, + "teaching_rms": 0.0009007894767279102, + "traffic_rms": 0.19860360287655857 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.50982865788319e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4563283095965293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005190200003308769, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.007033691170886159 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07101801037788391, + "mean_feedback_forward_cosine": 0.9994416063482111, + "mean_feedback_forward_relative_error": 0.031091597278348423, + "min_feedback_forward_cosine": 0.9974935054779053 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009061122772729981, + "instruction_rms": 0.000906112277275743, + "raw_apical_rms": 0.19862110934112187, + "teaching_rms": 0.0009061122772729981, + "traffic_rms": 0.198617491035837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.835217902789736e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4556975960006008e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005254594622289153, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.007410051720142364 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07100477814674377, + "mean_feedback_forward_cosine": 0.9994418675249274, + "mean_feedback_forward_relative_error": 0.03108482180501927, + "min_feedback_forward_cosine": 0.9974945187568665 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010888866943520256, + "instruction_rms": 0.0010888866945878372, + "raw_apical_rms": 0.1986044027406809, + "teaching_rms": 0.0010888866943520256, + "traffic_rms": 0.19859936956320898 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.409127616459955e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4541310576878738e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005194665828949503, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.009170027937218547 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07099320739507675, + "mean_feedback_forward_cosine": 0.999442085352811, + "mean_feedback_forward_relative_error": 0.03107914309948683, + "min_feedback_forward_cosine": 0.9974953532218933 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000984304165236274, + "instruction_rms": 0.000984304165384461, + "raw_apical_rms": 0.19859773397355449, + "teaching_rms": 0.000984304165236274, + "traffic_rms": 0.19859388161013422 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.223627480039795e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4547594358086225e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005138192616022023, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.008122862913832069 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07098197191953659, + "mean_feedback_forward_cosine": 0.9994423096830195, + "mean_feedback_forward_relative_error": 0.0310732192275199, + "min_feedback_forward_cosine": 0.9974960088729858 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009590896732649674, + "instruction_rms": 0.0009590896730412782, + "raw_apical_rms": 0.1985874397189068, + "teaching_rms": 0.0009590896732649674, + "traffic_rms": 0.19858342813541927 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.367846504597765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4584751411201764e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00517402420489567, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.008091802515983581 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07097038626670837, + "mean_feedback_forward_cosine": 0.9994425242597407, + "mean_feedback_forward_relative_error": 0.03106747240844098, + "min_feedback_forward_cosine": 0.9974969625473022 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009812022565942814, + "instruction_rms": 0.0009812022562432654, + "raw_apical_rms": 0.19859494733626848, + "teaching_rms": 0.0009812022565942814, + "traffic_rms": 0.19859097493718963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.37472305445408e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.460843900335787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005151572334256666, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.007736592751443386 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095889747142792, + "mean_feedback_forward_cosine": 0.9994427485899492, + "mean_feedback_forward_relative_error": 0.031061977427452803, + "min_feedback_forward_cosine": 0.9974977374076843 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000872387732203055, + "instruction_rms": 0.0008723877321707823, + "raw_apical_rms": 0.19864231295446036, + "teaching_rms": 0.000872387732203055, + "traffic_rms": 0.19863883515276773 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.5650190233791363e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.453675856800904e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005170748556815396, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.006516078158915043 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0709473118185997, + "mean_feedback_forward_cosine": 0.9994429414922541, + "mean_feedback_forward_relative_error": 0.031056620654734697, + "min_feedback_forward_cosine": 0.9974985718727112 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008759879590589954, + "instruction_rms": 0.0008759879588691847, + "raw_apical_rms": 0.19853133641943577, + "teaching_rms": 0.0008759879590589954, + "traffic_rms": 0.19852795888041933 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.389691264985231e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4568714794503277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005181208992771188, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006663693172633648 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07093457132577896, + "mean_feedback_forward_cosine": 0.999443192915483, + "mean_feedback_forward_relative_error": 0.031050480030138383, + "min_feedback_forward_cosine": 0.9974994659423828 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000917162858437985, + "instruction_rms": 0.0009171628581629316, + "raw_apical_rms": 0.19848226363117388, + "teaching_rms": 0.000917162858437985, + "traffic_rms": 0.19847836122062418 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.4883622858124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4604533689798454e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051666042424135085, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.006982537309993059 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07092170417308807, + "mean_feedback_forward_cosine": 0.9994434020736, + "mean_feedback_forward_relative_error": 0.03104492685842243, + "min_feedback_forward_cosine": 0.9975001811981201 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008657565970352031, + "instruction_rms": 0.000865756596617077, + "raw_apical_rms": 0.1985229202868265, + "teaching_rms": 0.0008657565970352031, + "traffic_rms": 0.1985194211660336 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.942541202510256e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.462424831445963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00523605273910061, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006335549934282899 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091014832258224, + "mean_feedback_forward_cosine": 0.9994436502456665, + "mean_feedback_forward_relative_error": 0.03103921368548816, + "min_feedback_forward_cosine": 0.9975011348724365 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000860342801093788, + "instruction_rms": 0.0008603428012837423, + "raw_apical_rms": 0.19844662791649098, + "teaching_rms": 0.000860342801093788, + "traffic_rms": 0.19844310960572917 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.447817213484086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4560085214198357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005207154316572451, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.006344442099118605 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07089883089065552, + "mean_feedback_forward_cosine": 0.9994438518177379, + "mean_feedback_forward_relative_error": 0.03103379345583645, + "min_feedback_forward_cosine": 0.9975018501281738 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008382857888126977, + "instruction_rms": 0.0008382857892124851, + "raw_apical_rms": 0.19840592761389877, + "teaching_rms": 0.0008382857888126977, + "traffic_rms": 0.19840255258273826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.832058119243811e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.457895069551827e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005190139738145289, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.006669815689213574 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088804990053177, + "mean_feedback_forward_cosine": 0.9994440501386469, + "mean_feedback_forward_relative_error": 0.031028075465424493, + "min_feedback_forward_cosine": 0.9975026249885559 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008600565823382211, + "instruction_rms": 0.0008600565821161378, + "raw_apical_rms": 0.19841534225786578, + "teaching_rms": 0.0008600565823382211, + "traffic_rms": 0.19841198383300407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.328612383616928e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4514694501456336e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005274943251047321, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.0065101036351174115 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087628543376923, + "mean_feedback_forward_cosine": 0.9994442744688554, + "mean_feedback_forward_relative_error": 0.03102250546216965, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008405675066713958, + "instruction_rms": 0.0008405675066287487, + "raw_apical_rms": 0.1984425367431237, + "teaching_rms": 0.0008405675066713958, + "traffic_rms": 0.1984392261301603 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.704645099693153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4577687555376206e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005199751495553181, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005932280392274261 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087613642215729, + "mean_feedback_forward_cosine": 0.9994442723014137, + "mean_feedback_forward_relative_error": 0.031022419484162872, + "min_feedback_forward_cosine": 0.9975034594535828 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008352126211054606, + "instruction_rms": 0.0008352126207653732, + "raw_apical_rms": 0.19843134541717114, + "teaching_rms": 0.0008352126211054606, + "traffic_rms": 0.1984280325597218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.00494879879443e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4568140284079475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005184273900987599, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005908031362853944 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087605446577072, + "mean_feedback_forward_cosine": 0.9994442690502513, + "mean_feedback_forward_relative_error": 0.0310223570516841, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006999618465159179, + "instruction_rms": 0.000699961846543442, + "raw_apical_rms": 0.1984319939331712, + "teaching_rms": 0.0006999618465159179, + "traffic_rms": 0.19842965797833936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.586440631257574e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4565748106495767e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005188715872578883, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.005021687182933092 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087596505880356, + "mean_feedback_forward_cosine": 0.9994442907246677, + "mean_feedback_forward_relative_error": 0.03102229774336923, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006653337246858778, + "instruction_rms": 0.000665333724778397, + "raw_apical_rms": 0.19843014664425418, + "teaching_rms": 0.0006653337246858778, + "traffic_rms": 0.19842783767475164 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.923963979943437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4558876165487727e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005227421194972849, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0046188938918570056 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087589055299759, + "mean_feedback_forward_cosine": 0.9994442809711803, + "mean_feedback_forward_relative_error": 0.031022238697518002, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006755021573545577, + "instruction_rms": 0.0006755021570261319, + "raw_apical_rms": 0.19842687383434635, + "teaching_rms": 0.0006755021573545577, + "traffic_rms": 0.19842464889854403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.500596327896346e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4544007476849943e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005186066573474382, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004739633579477668 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087580859661102, + "mean_feedback_forward_cosine": 0.9994442950595509, + "mean_feedback_forward_relative_error": 0.03102217670530081, + "min_feedback_forward_cosine": 0.9975034594535828 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000689985292663778, + "instruction_rms": 0.0006899852929873441, + "raw_apical_rms": 0.19842403339644812, + "teaching_rms": 0.000689985292663778, + "traffic_rms": 0.1984216114021958 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 7.766475818016261e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.458719488706358e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005169648083257179, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004798793706316501 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087571173906326, + "mean_feedback_forward_cosine": 0.9994442863897843, + "mean_feedback_forward_relative_error": 0.031022117549384182, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007127411481667645, + "instruction_rms": 0.0007127411479346683, + "raw_apical_rms": 0.19841267249078187, + "teaching_rms": 0.0007127411481667645, + "traffic_rms": 0.19840994500359008 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.130058238795755e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4588388363485124e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052111710106986795, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004737346856324002 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708756148815155, + "mean_feedback_forward_cosine": 0.9994442798874595, + "mean_feedback_forward_relative_error": 0.031022056183693084, + "min_feedback_forward_cosine": 0.9975034594535828 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007279549564201155, + "instruction_rms": 0.0007279549563313103, + "raw_apical_rms": 0.19840619771315238, + "teaching_rms": 0.0007279549564201155, + "traffic_rms": 0.19840366162333237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.810090980025052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4560076263865057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005198181196058761, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004792956570647657 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087555527687073, + "mean_feedback_forward_cosine": 0.9994442874735052, + "mean_feedback_forward_relative_error": 0.031022003589367325, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000628039039688138, + "instruction_rms": 0.0006280390396776572, + "raw_apical_rms": 0.19840040043594118, + "teaching_rms": 0.000628039039688138, + "traffic_rms": 0.19839849235125653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.354715592100547e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4546408912466624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005201924671441793, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0041045606086030605 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087548077106476, + "mean_feedback_forward_cosine": 0.9994442939758301, + "mean_feedback_forward_relative_error": 0.031021949377926912, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006422627345033392, + "instruction_rms": 0.0006422627346204936, + "raw_apical_rms": 0.19840430545017262, + "teaching_rms": 0.0006422627345033392, + "traffic_rms": 0.1984022302216301 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.65345956715646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.456912295612518e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005212250307062712, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.004289394716303796 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070875383913517, + "mean_feedback_forward_cosine": 0.9994442918083885, + "mean_feedback_forward_relative_error": 0.031021888503296808, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006664954500301222, + "instruction_rms": 0.0006664954498556702, + "raw_apical_rms": 0.19840809729194273, + "teaching_rms": 0.0006664954500301222, + "traffic_rms": 0.19840546908413345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.301002031250391e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4570145048919254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005235190130952043, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.004854296311326325 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087531685829163, + "mean_feedback_forward_cosine": 0.9994442928921092, + "mean_feedback_forward_relative_error": 0.031021835045381026, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000583682077381, + "instruction_rms": 0.0005836820777246466, + "raw_apical_rms": 0.19840586554877504, + "teaching_rms": 0.000583682077381, + "traffic_rms": 0.19840401687219067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.867010699489583e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.463414015475103e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005159691509365488, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004024033974134363 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087524980306625, + "mean_feedback_forward_cosine": 0.9994442993944341, + "mean_feedback_forward_relative_error": 0.031021780554543844, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006109186453110178, + "instruction_rms": 0.0006109186451682414, + "raw_apical_rms": 0.19839504249994439, + "teaching_rms": 0.0006109186453110178, + "traffic_rms": 0.1983929002396999 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.151612552685037e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4579942197758857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005201297746984306, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.003951676387898624 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087517529726028, + "mean_feedback_forward_cosine": 0.9994442918083885, + "mean_feedback_forward_relative_error": 0.03102173221890222, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00059490662006452, + "instruction_rms": 0.0005949066203547712, + "raw_apical_rms": 0.19839937424639822, + "teaching_rms": 0.00059490662006452, + "traffic_rms": 0.19839749698010006 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.013488811935531e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4570330587987505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005213150061505643, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.003897927967980504 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708751231431961, + "mean_feedback_forward_cosine": 0.9994443048130383, + "mean_feedback_forward_relative_error": 0.031021677787331017, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006071230650155615, + "instruction_rms": 0.0006071230649386183, + "raw_apical_rms": 0.19840685310204723, + "teaching_rms": 0.0006071230650155615, + "traffic_rms": 0.19840477093026393 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 7.434940130224277e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4529379590278897e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005212620588886794, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.0038895293265208603 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087503373622894, + "mean_feedback_forward_cosine": 0.9994442961432717, + "mean_feedback_forward_relative_error": 0.031021623431958936, + "min_feedback_forward_cosine": 0.9975034594535828 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006038273426023746, + "instruction_rms": 0.0006038273425548478, + "raw_apical_rms": 0.1983977096088115, + "teaching_rms": 0.0006038273426023746, + "traffic_rms": 0.1983956456995935 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.574558852889822e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4550262182414503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005162958637371182, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.004045673833340407 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087497413158417, + "mean_feedback_forward_cosine": 0.9994443048130383, + "mean_feedback_forward_relative_error": 0.03102157186208801, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005856811854687501, + "instruction_rms": 0.0005856811851808009, + "raw_apical_rms": 0.19839649014247737, + "teaching_rms": 0.0005856811854687501, + "traffic_rms": 0.1983945624002354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1365042951183568e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4548727531945382e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005175048068519982, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0037456664326414465 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087491452693939, + "mean_feedback_forward_cosine": 0.9994443004781549, + "mean_feedback_forward_relative_error": 0.0310215188190341, + "min_feedback_forward_cosine": 0.997503399848938 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006099395495430482, + "instruction_rms": 0.0006099395492629795, + "raw_apical_rms": 0.19839931493149326, + "teaching_rms": 0.0006099395495430482, + "traffic_rms": 0.19839731691280715 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.263616825208373e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4635543236011635e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005256844875537188, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003853200528845191 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087483257055283, + "mean_feedback_forward_cosine": 0.9994443145665255, + "mean_feedback_forward_relative_error": 0.03102146831595085, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005324029095230174, + "instruction_rms": 0.0005324029096370997, + "raw_apical_rms": 0.19840162015261154, + "teaching_rms": 0.0005324029095230174, + "traffic_rms": 0.1983999675930895 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.490534414955619e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.454272403742583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005212492343790001, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.0035581306520663203 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087477296590805, + "mean_feedback_forward_cosine": 0.9994443069804798, + "mean_feedback_forward_relative_error": 0.031021423637866974, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005139249229244815, + "instruction_rms": 0.0005139249230309824, + "raw_apical_rms": 0.19840020150583032, + "teaching_rms": 0.0005139249229244815, + "traffic_rms": 0.19839867351053583 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.274450553040879e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4592641982005717e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005204792962792407, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0033015960488282144 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087469846010208, + "mean_feedback_forward_cosine": 0.9994443199851296, + "mean_feedback_forward_relative_error": 0.031021371043541215, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005931269877756821, + "instruction_rms": 0.0005931269875862437, + "raw_apical_rms": 0.19840348602524016, + "teaching_rms": 0.0005931269877756821, + "traffic_rms": 0.19840147471450945 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.369310850435795e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4610529369226817e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051939545537208545, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0038284460739418866 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087463140487671, + "mean_feedback_forward_cosine": 0.9994443123990839, + "mean_feedback_forward_relative_error": 0.031021322013640946, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005691439783579176, + "instruction_rms": 0.0005691439781327346, + "raw_apical_rms": 0.19838013811621222, + "teaching_rms": 0.0005691439783579176, + "traffic_rms": 0.198378199937522 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.042215700408633e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4548119881957364e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00517076741808568, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0034550998530164362 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087455689907074, + "mean_feedback_forward_cosine": 0.9994443026455966, + "mean_feedback_forward_relative_error": 0.031021267514337193, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00060687613509259, + "instruction_rms": 0.000606876135179811, + "raw_apical_rms": 0.19838131946574805, + "teaching_rms": 0.00060687613509259, + "traffic_rms": 0.19837943015695114 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.362449767038925e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4533974390634368e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052044932984464986, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.004053802644014359 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087449729442596, + "mean_feedback_forward_cosine": 0.9994443221525713, + "mean_feedback_forward_relative_error": 0.031021217375316403, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005797225152001788, + "instruction_rms": 0.0005797225152039247, + "raw_apical_rms": 0.19837943820325557, + "teaching_rms": 0.0005797225152001788, + "traffic_rms": 0.19837769694143376 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.384224349176293e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.453052132987642e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005231493064910818, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0036838062054291368 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087443023920059, + "mean_feedback_forward_cosine": 0.9994443080642007, + "mean_feedback_forward_relative_error": 0.03102117201990702, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005703037511427691, + "instruction_rms": 0.0005703037510203516, + "raw_apical_rms": 0.19837876783247854, + "teaching_rms": 0.0005703037511427691, + "traffic_rms": 0.19837708829018713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 5.831750371498856e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4511705868657675e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005264586799461567, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.00366726968402043 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087436318397522, + "mean_feedback_forward_cosine": 0.9994443123990839, + "mean_feedback_forward_relative_error": 0.031021125614643097, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005439385730983528, + "instruction_rms": 0.0005439385732534964, + "raw_apical_rms": 0.19837582204999543, + "teaching_rms": 0.0005439385730983528, + "traffic_rms": 0.19837412421842746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.28502476729409e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4543267350132903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052337591668257626, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.003257799834087491 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087429612874985, + "mean_feedback_forward_cosine": 0.9994443048130383, + "mean_feedback_forward_relative_error": 0.031021074298769234, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005551257525163542, + "instruction_rms": 0.0005551257526217705, + "raw_apical_rms": 0.19838754615142104, + "teaching_rms": 0.0005551257525163542, + "traffic_rms": 0.19838574621616303 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.116390404211415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4519769907809278e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00525964115245736, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0034607375524565577 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087421417236328, + "mean_feedback_forward_cosine": 0.9994443210688504, + "mean_feedback_forward_relative_error": 0.031021020307459613, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005570930639744078, + "instruction_rms": 0.0005570930638937304, + "raw_apical_rms": 0.19839688205822775, + "teaching_rms": 0.0005570930639744078, + "traffic_rms": 0.19839502020766048 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 7.441861527013316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4626235941380288e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00524380928434035, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0034764549234882 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708741694688797, + "mean_feedback_forward_cosine": 0.9994443026455966, + "mean_feedback_forward_relative_error": 0.03102098039605401, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000487896554721914, + "instruction_rms": 0.00048789655472409736, + "raw_apical_rms": 0.19838515423679295, + "teaching_rms": 0.000487896554721914, + "traffic_rms": 0.19838381448873826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.450164136140302e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.455255612615303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005170866890442281, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0030614467611908914 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087410241365433, + "mean_feedback_forward_cosine": 0.9994443254037337, + "mean_feedback_forward_relative_error": 0.031020933855324985, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005256534224849093, + "instruction_rms": 0.000525653422310863, + "raw_apical_rms": 0.1983885388303093, + "teaching_rms": 0.0005256534224849093, + "traffic_rms": 0.1983868748011363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.060152202280733e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4559721261704332e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005193507642783886, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.003334190772008151 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087403535842896, + "mean_feedback_forward_cosine": 0.9994443178176879, + "mean_feedback_forward_relative_error": 0.031020883530039678, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005949596379681088, + "instruction_rms": 0.0005949596378681266, + "raw_apical_rms": 0.19839655449087715, + "teaching_rms": 0.0005949596379681088, + "traffic_rms": 0.19839459661691453 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.813817859414485e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4605804579684185e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005341173906741259, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0039058992750942707 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087396830320358, + "mean_feedback_forward_cosine": 0.9994443221525713, + "mean_feedback_forward_relative_error": 0.03102083448320627, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005557876442186892, + "instruction_rms": 0.0005557876442407051, + "raw_apical_rms": 0.19838762858299583, + "teaching_rms": 0.0005557876442186892, + "traffic_rms": 0.19838569347218332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.031283644209907e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4622970568169133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052170114473527275, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0037654544898681344 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087390124797821, + "mean_feedback_forward_cosine": 0.9994443189014088, + "mean_feedback_forward_relative_error": 0.03102078405632214, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005761374741845923, + "instruction_rms": 0.0005761374738444317, + "raw_apical_rms": 0.19836357234668672, + "teaching_rms": 0.0005761374741845923, + "traffic_rms": 0.1983617425354422 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.807526714032065e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4539504648635845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005172793975464787, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0034998789017274975 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087381184101105, + "mean_feedback_forward_cosine": 0.9994443275711753, + "mean_feedback_forward_relative_error": 0.031020724138414318, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006179242252465211, + "instruction_rms": 0.0006179242253167938, + "raw_apical_rms": 0.19837288702480932, + "teaching_rms": 0.0006179242252465211, + "traffic_rms": 0.1983705226275133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.173841882173292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.45340419472121e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005222758028199636, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.004100104279592633 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087374478578568, + "mean_feedback_forward_cosine": 0.9994443243200128, + "mean_feedback_forward_relative_error": 0.03102067204361612, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005867640849503622, + "instruction_rms": 0.0005867640847766611, + "raw_apical_rms": 0.19837360630061016, + "teaching_rms": 0.0005867640849503622, + "traffic_rms": 0.19837165266139067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 8.053980593558663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.45679067698397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005223032633933359, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0038313682655058802 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708736702799797, + "mean_feedback_forward_cosine": 0.9994443329897794, + "mean_feedback_forward_relative_error": 0.031020619271492415, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005745299370221307, + "instruction_rms": 0.0005745299368285423, + "raw_apical_rms": 0.19837839507076455, + "teaching_rms": 0.0005745299370221307, + "traffic_rms": 0.19837643945232725 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.368669313085775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4550576538034435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005175468355130116, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0036198436260037126 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087359577417374, + "mean_feedback_forward_cosine": 0.9994443373246626, + "mean_feedback_forward_relative_error": 0.031020566753365775, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000569675066957006, + "instruction_rms": 0.0005696750671878628, + "raw_apical_rms": 0.19837954847705666, + "teaching_rms": 0.000569675066957006, + "traffic_rms": 0.19837768878964848 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 7.080065245190781e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4580503526578067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005212411001893131, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0035563668692857025 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087353616952896, + "mean_feedback_forward_cosine": 0.9994443308223377, + "mean_feedback_forward_relative_error": 0.031020517308603633, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005192432622622501, + "instruction_rms": 0.0005192432622952186, + "raw_apical_rms": 0.19837143116241368, + "teaching_rms": 0.0005192432622622501, + "traffic_rms": 0.1983698686461472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.119332678939827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.45699557397128e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051989044141162845, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0032445984718203545 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087347656488419, + "mean_feedback_forward_cosine": 0.9994443319060585, + "mean_feedback_forward_relative_error": 0.031020470852540297, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005174913985374893, + "instruction_rms": 0.0005174913985395044, + "raw_apical_rms": 0.19835766008898922, + "teaching_rms": 0.0005174913985374893, + "traffic_rms": 0.19835612523492654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 8.423488395692402e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4542211095288394e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005195300166156281, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003139071330465376 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087341696023941, + "mean_feedback_forward_cosine": 0.9994443329897794, + "mean_feedback_forward_relative_error": 0.031020428062501278, + "min_feedback_forward_cosine": 0.9975035786628723 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005124451325271137, + "instruction_rms": 0.0005124451323233437, + "raw_apical_rms": 0.19836225822850184, + "teaching_rms": 0.0005124451325271137, + "traffic_rms": 0.19836083591956438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 7.309558469614785e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.455061846853004e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0051700029931921136, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003097061588875949 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087335735559464, + "mean_feedback_forward_cosine": 0.9994443340735002, + "mean_feedback_forward_relative_error": 0.031020381860435008, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005216481368464296, + "instruction_rms": 0.0005216481367170813, + "raw_apical_rms": 0.1983520140883415, + "teaching_rms": 0.0005216481368464296, + "traffic_rms": 0.1983504653078767 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679288148880005, + "maximum_absolute_post_projection_soma_slope": 6.230025917375315e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.455474367166569e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005260110822380021, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0031969439377449453 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087327539920807, + "mean_feedback_forward_cosine": 0.9994443319060585, + "mean_feedback_forward_relative_error": 0.031020332441072572, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005436299421096169, + "instruction_rms": 0.0005436299419663652, + "raw_apical_rms": 0.1983496793655065, + "teaching_rms": 0.0005436299421096169, + "traffic_rms": 0.1983479009987772 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679291129112244, + "maximum_absolute_post_projection_soma_slope": 6.450100897836819e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4587179143368043e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052313750355796685, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0032822404459211976 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087323814630508, + "mean_feedback_forward_cosine": 0.9994443340735002, + "mean_feedback_forward_relative_error": 0.031020289507101884, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004764600348086196, + "instruction_rms": 0.00047646003455047094, + "raw_apical_rms": 0.19834649462170426, + "teaching_rms": 0.0004764600348086196, + "traffic_rms": 0.19834510507460187 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.795836782202969e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4571782095662998e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005224601649212932, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0030518958858028055 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087316364049911, + "mean_feedback_forward_cosine": 0.9994443264874545, + "mean_feedback_forward_relative_error": 0.0310202428732406, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005438428453949378, + "instruction_rms": 0.0005438428454955388, + "raw_apical_rms": 0.19834393393682032, + "teaching_rms": 0.0005438428453949378, + "traffic_rms": 0.19834217649827485 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.260765665430881e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4575738280612573e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005233951139437571, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0033357055381685497 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087308168411255, + "mean_feedback_forward_cosine": 0.99944434707815, + "mean_feedback_forward_relative_error": 0.03102018926292658, + "min_feedback_forward_cosine": 0.9975037574768066 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005745461595731847, + "instruction_rms": 0.0005745461592588483, + "raw_apical_rms": 0.19834217799562165, + "teaching_rms": 0.0005745461595731847, + "traffic_rms": 0.1983402787520629 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.444171501447272e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4548498662341882e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005200864648099261, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0035722616494633258 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087302207946777, + "mean_feedback_forward_cosine": 0.9994443319060585, + "mean_feedback_forward_relative_error": 0.031020138192583212, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005338283003820097, + "instruction_rms": 0.0005338283003500636, + "raw_apical_rms": 0.19834741290401714, + "teaching_rms": 0.0005338283003820097, + "traffic_rms": 0.1983456250202075 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 1.1837356339583494e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4628018867091537e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052015247819041715, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0031789246093854306 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708729475736618, + "mean_feedback_forward_cosine": 0.9994443416595459, + "mean_feedback_forward_relative_error": 0.031020083523948083, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005852475355660669, + "instruction_rms": 0.0005852475355995258, + "raw_apical_rms": 0.1983580408783285, + "teaching_rms": 0.0005852475355660669, + "traffic_rms": 0.19835593561391532 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.345303032479933e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4559123164556096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0052775150593144865, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0036035204985551536 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087288796901703, + "mean_feedback_forward_cosine": 0.9994443286548961, + "mean_feedback_forward_relative_error": 0.03102003690701994, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005302380234393901, + "instruction_rms": 0.0005302380232493751, + "raw_apical_rms": 0.1983530837425426, + "teaching_rms": 0.0005302380234393901, + "traffic_rms": 0.19835147411459012 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.972496973707166e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4579206035445845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005249619023913873, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.003161972364485264 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087282091379166, + "mean_feedback_forward_cosine": 0.9994443427432668, + "mean_feedback_forward_relative_error": 0.031019986446269535, + "min_feedback_forward_cosine": 0.9975036382675171 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005336209022694507, + "instruction_rms": 0.0005336209024587929, + "raw_apical_rms": 0.19836216862885736, + "teaching_rms": 0.0005336209022694507, + "traffic_rms": 0.1983604383334117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 9.765429354047228e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4551876547590628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005240751925353194, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.003450669647715986 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087274640798569, + "mean_feedback_forward_cosine": 0.9994443514130332, + "mean_feedback_forward_relative_error": 0.031019937026907096, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005496137953538349, + "instruction_rms": 0.0005496137951772304, + "raw_apical_rms": 0.19837051343171147, + "teaching_rms": 0.0005496137953538349, + "traffic_rms": 0.1983686675752752 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 6.517190342947288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4576794100733942e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005266707241306954, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0033774491391144693 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087264955043793, + "mean_feedback_forward_cosine": 0.9994443373246626, + "mean_feedback_forward_relative_error": 0.031019884212450548, + "min_feedback_forward_cosine": 0.9975036978721619 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005897003023443545, + "instruction_rms": 0.0005897003024692283, + "raw_apical_rms": 0.19836524352767781, + "teaching_rms": 0.0005897003023443545, + "traffic_rms": 0.19836318098523031 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.21679289638996124, + "maximum_absolute_post_projection_soma_slope": 7.27837843328416e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.457650115476823e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005154803657624553, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0036807414781674743 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9292, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3912335693359375, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3324800000, + "peak_memory_reserved_bytes": 5553258496, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 1.2472722232814704e-07, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 1.2146459482664795e-07, + "min_residual_soma_slope": -1.2472722232814704e-07, + "mse": 1.8070805926959287e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.199251019882141e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.199251019882141e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 10, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0234105587005615, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.06387972831726074, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 26448.92703437805, + "train_wall_s": 26442.46849989891 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.056737203150987625, + 0.03466915711760521, + 0.025380922481417656, + 0.023368151858448982, + 0.015401496551930904, + 0.01708696223795414, + 0.010191723704338074, + 0.012498791329562664, + 0.007734539452940226, + 0.009953099302947521, + 0.006193092558532953, + 0.008852911181747913, + 0.004943703766912222, + 0.006862254347652197, + 0.004266873002052307, + 0.006199307274073362, + 0.003704395378008485, + 0.005341028328984976, + 0.0031979475170373917, + 0.007326237857341766, + 0.005027200561016798, + 0.006448567379266024, + 0.00441219424828887, + 0.006279183551669121, + 0.003818117780610919, + 0.005211378913372755, + 0.003340089460834861, + 0.004735496826469898, + 0.003009872045367956, + 0.004164178390055895, + 0.0027639002073556185, + 0.003919674549251795, + 0.0024971873499453068, + 0.0037482266779989004, + 0.0022654798813164234, + 0.0033104282338172197, + 0.0021054078824818134, + 0.004926503635942936, + 0.003475334495306015, + 0.004802375566214323, + 0.0031876815482974052, + 0.004390588030219078, + 0.0029604053124785423, + 0.004102425649762154, + 0.0027928128838539124, + 0.00390723068267107, + 0.002610700437799096, + 0.0035815658047795296, + 0.0025039466563612223, + 0.003361599752679467, + 0.0024003200232982635, + 0.0032344518695026636, + 0.002335174474865198, + 0.003047503065317869, + 0.002387239597737789 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.000000476837158, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.30167168378829956, + 0.18403391540050507, + 0.09050484746694565, + 0.1256996989250183, + 0.04250982776284218, + 0.09509823471307755, + 0.023601505905389786, + 0.06689246743917465, + 0.016166171059012413, + 0.05356940999627113, + 0.011292538605630398, + 0.047968730330467224, + 0.008278507739305496, + 0.03684104233980179, + 0.006637901067733765, + 0.03280358761548996, + 0.005378217436373234, + 0.028896356001496315, + 0.004386315122246742, + 0.03771384805440903, + 0.00903933122754097, + 0.03365190327167511, + 0.007395597640424967, + 0.03414306789636612, + 0.005943499505519867, + 0.027080019935965538, + 0.004866709932684898, + 0.02507423609495163, + 0.004125508479773998, + 0.021414360031485558, + 0.003599257906898856, + 0.020058082416653633, + 0.0031137375626713037, + 0.019754327833652496, + 0.0026740720495581627, + 0.017418887466192245, + 0.002392395166680217, + 0.025816624984145164, + 0.005350381135940552, + 0.02607998624444008, + 0.004654225427657366, + 0.022968623787164688, + 0.004061016254127026, + 0.021902546286582947, + 0.0036145965568721294, + 0.02107754908502102, + 0.0031958860345184803, + 0.01884119212627411, + 0.002985289553180337, + 0.018127063289284706, + 0.0027052839286625385, + 0.017298173159360886, + 0.0025729695335030556, + 0.01592264696955681, + 0.002482388401404023 + ], + "unscaled_traffic_rms": [ + 0.7523040175437927, + 0.7535384297370911, + 1.121748685836792, + 0.743618369102478, + 1.449217438697815, + 0.7187078595161438, + 1.7273005247116089, + 0.7473960518836975, + 1.9137591123580933, + 0.743192732334137, + 2.1936936378479004, + 0.7382234930992126, + 2.38869309425354, + 0.745066225528717, + 2.571218252182007, + 0.7559303641319275, + 2.755110263824463, + 0.7393358945846558, + 2.916295289993286, + 0.7770342230796814, + 2.2245895862579346, + 0.7665025591850281, + 2.3863894939422607, + 0.735632061958313, + 2.5696091651916504, + 0.7697747349739075, + 2.7452545166015625, + 0.7554362416267395, + 2.918303966522217, + 0.7778291702270508, + 3.0716333389282227, + 0.7816648483276367, + 3.207961320877075, + 0.7589682340621948, + 3.3888089656829834, + 0.7601928114891052, + 3.520167589187622, + 0.7633071541786194, + 2.598195791244507, + 0.7365610599517822, + 2.7396023273468018, + 0.7646236419677734, + 2.9159257411956787, + 0.7492143511772156, + 3.0905942916870117, + 0.741496205329895, + 3.2675764560699463, + 0.7603692412376404, + 3.3550469875335693, + 0.7417858242988586, + 3.549084186553955, + 0.7479291558265686, + 3.6303181648254395, + 0.7655770182609558, + 3.846681833267212 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250441282772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 127796290519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 5010596513857536 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json new file mode 100644 index 0000000..149dc20 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json @@ -0,0 +1,7293 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 11, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 11, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5011, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9994528922769759, + "feedback_forward_cosine": [ + 0.9996585845947266, + 0.9996613264083862, + 0.9994030594825745, + 0.9993560910224915, + 0.9991961121559143, + 0.9991893768310547, + 0.9993095397949219, + 0.9992635250091553, + 0.9995965361595154, + 0.9995520114898682, + 0.9995437860488892, + 0.9994943141937256, + 0.9995837211608887, + 0.99957674741745, + 0.999679684638977, + 0.999646782875061, + 0.9997938871383667, + 0.9996997714042664, + 0.9997414350509644, + 0.9998083114624023, + 0.9996501803398132, + 0.9995943307876587, + 0.9997068047523499, + 0.9996200799942017, + 0.9997164011001587, + 0.9996328353881836, + 0.9996664524078369, + 0.9995362758636475, + 0.9996434450149536, + 0.9995038509368896, + 0.9996674656867981, + 0.9994937181472778, + 0.9995723962783813, + 0.9994343519210815, + 0.9995824098587036, + 0.9993937611579895, + 0.9997337460517883, + 0.9998224377632141, + 0.9997802972793579, + 0.9997556209564209, + 0.9997716546058655, + 0.999693751335144, + 0.9997555017471313, + 0.999632716178894, + 0.9997289776802063, + 0.9995256662368774, + 0.9993695020675659, + 0.9991015195846558, + 0.999431312084198, + 0.9991052746772766, + 0.9991822838783264, + 0.9987474679946899, + 0.9989230036735535, + 0.9982337951660156, + 0.9999802112579346 + ], + "feedback_forward_norm_ratio": [ + 0.9989251494407654, + 1.000936508178711, + 0.9989844560623169, + 1.001094937324524, + 0.9990232586860657, + 1.0019923448562622, + 0.99794602394104, + 1.001634120941162, + 1.0001002550125122, + 1.0009591579437256, + 0.9991543889045715, + 1.002602219581604, + 0.9991486072540283, + 1.00127375125885, + 1.0000529289245605, + 1.0020766258239746, + 0.9994543194770813, + 1.000896692276001, + 0.9993245005607605, + 1.0006946325302124, + 0.9990779757499695, + 1.001394271850586, + 0.9992592334747314, + 1.0014909505844116, + 0.9996901154518127, + 1.0012389421463013, + 0.9994320273399353, + 1.0017030239105225, + 0.9988840222358704, + 1.001800298690796, + 0.9995790719985962, + 1.0016040802001953, + 0.9988933205604553, + 1.0019941329956055, + 0.9989625811576843, + 1.001695990562439, + 0.9992861747741699, + 1.0003782510757446, + 0.9997970461845398, + 1.0006024837493896, + 0.9996731281280518, + 1.0012645721435547, + 0.9997552633285522, + 1.001225233078003, + 0.9995972514152527, + 1.0016262531280518, + 0.9985372424125671, + 1.0021247863769531, + 0.9986729621887207, + 1.0028369426727295, + 0.9981772303581238, + 1.00296151638031, + 0.9981222152709961, + 1.002724051475525, + 1.0013223886489868 + ], + "feedback_forward_relative_error": [ + 0.026138870045542717, + 0.02605477347970009, + 0.03454797342419624, + 0.03592420741915703, + 0.040088262408971786, + 0.040355466306209564, + 0.037179701030254364, + 0.03844698145985603, + 0.028409859165549278, + 0.02996210940182209, + 0.03020460158586502, + 0.0319485142827034, + 0.028850506991147995, + 0.029142992570996284, + 0.02531236596405506, + 0.02668643370270729, + 0.020307406783103943, + 0.02453530579805374, + 0.022739969193935394, + 0.019597264006733894, + 0.02645433507859707, + 0.02853545919060707, + 0.02421659789979458, + 0.027628593146800995, + 0.02381478250026703, + 0.027145279571413994, + 0.02582474984228611, + 0.03052673302590847, + 0.026714729145169258, + 0.031582459807395935, + 0.02578607387840748, + 0.03188617154955864, + 0.029251690953969955, + 0.03372754156589508, + 0.028902731835842133, + 0.034893520176410675, + 0.023077230900526047, + 0.018855642527341843, + 0.020960478112101555, + 0.022119291126728058, + 0.021369803696870804, + 0.024794884026050568, + 0.022112904116511345, + 0.02714836597442627, + 0.023276351392269135, + 0.030870573595166206, + 0.0355137400329113, + 0.04249107465147972, + 0.033728212118148804, + 0.04245596006512642, + 0.04044438898563385, + 0.05021299421787262, + 0.046406086534261703, + 0.05957596004009247, + 0.006421781610697508 + ], + "innovation_negative_gradient_cosine": [ + 0.9995517730712891, + 0.9994828701019287, + 0.9995512962341309, + 0.9991498589515686, + 0.9995520114898682, + 0.9991917610168457, + 0.9995540380477905, + 0.9992571473121643, + 0.9995548725128174, + 0.9994254112243652, + 0.9995452761650085, + 0.9993563294410706, + 0.9995415210723877, + 0.9993796348571777, + 0.9995396733283997, + 0.9994672536849976, + 0.9995462894439697, + 0.9995050430297852, + 0.9995566606521606, + 0.9996000528335571, + 0.9995860457420349, + 0.9994322061538696, + 0.999593198299408, + 0.9994602203369141, + 0.9996090531349182, + 0.9994994401931763, + 0.9996220469474792, + 0.999465823173523, + 0.9996377825737, + 0.9994761943817139, + 0.9996447563171387, + 0.9994639158248901, + 0.9996585249900818, + 0.9994835257530212, + 0.9996689558029175, + 0.9994775652885437, + 0.9996813535690308, + 0.9997172951698303, + 0.9997225999832153, + 0.9996993541717529, + 0.9997093677520752, + 0.9996503591537476, + 0.9997342228889465, + 0.9997210502624512, + 0.9997873306274414, + 0.9998279809951782, + 0.9997203946113586, + 0.9997026920318604, + 0.9997576475143433, + 0.9997735023498535, + 0.9997548460960388, + 0.9996984601020813, + 0.9998241066932678, + 0.9998534321784973, + 0.6879556775093079 + ], + "instruction_negative_gradient_cosine": [ + 0.9995517730712891, + 0.9994828701019287, + 0.9995514154434204, + 0.9991498589515686, + 0.9995520114898682, + 0.9991917610168457, + 0.9995540380477905, + 0.9992572069168091, + 0.9995548725128174, + 0.9994254112243652, + 0.9995452761650085, + 0.9993562698364258, + 0.9995415210723877, + 0.9993797540664673, + 0.9995396137237549, + 0.9994672536849976, + 0.9995462894439697, + 0.9995050430297852, + 0.9995566606521606, + 0.9996000528335571, + 0.9995861053466797, + 0.9994322061538696, + 0.999593198299408, + 0.9994602203369141, + 0.9996090531349182, + 0.9994994401931763, + 0.9996219873428345, + 0.999465823173523, + 0.9996377825737, + 0.9994760751724243, + 0.9996446967124939, + 0.9994639158248901, + 0.9996585845947266, + 0.9994834661483765, + 0.9996689558029175, + 0.9994775652885437, + 0.9996813535690308, + 0.9997173547744751, + 0.9997225999832153, + 0.9996993541717529, + 0.9997093677520752, + 0.9996503591537476, + 0.9997342228889465, + 0.9997210502624512, + 0.9997873902320862, + 0.999828040599823, + 0.9997203946113586, + 0.9997027516365051, + 0.9997576475143433, + 0.9997735023498535, + 0.9997549057006836, + 0.9996984601020813, + 0.999824047088623, + 0.9998534917831421, + 0.6888552904129028 + ], + "matched_negative_gradient_cosine": [ + -0.00474204309284687, + -0.009176000952720642, + -0.004329458810389042, + -0.0001360562164336443, + -0.00134295504540205, + -0.0058915987610816956, + 0.0028759310953319073, + 0.002320106141269207, + 0.001732651493512094, + -0.010575078427791595, + 0.0007172836922109127, + -0.03010343573987484, + 0.008269624784588814, + 0.0316404290497303, + 0.006767967715859413, + -0.011222892440855503, + 0.013338076882064342, + 0.020049583166837692, + -0.0010996179189532995, + 0.02430870570242405, + 0.004063806962221861, + 0.03761480748653412, + -0.005386119708418846, + 0.010938838124275208, + -0.011171958409249783, + 0.038992464542388916, + -0.021670546382665634, + -0.007791977375745773, + -0.013821092434227467, + -0.020795801654458046, + -0.00878394115716219, + 0.024180587381124496, + -0.007901962846517563, + 0.02275543287396431, + -0.012234559282660484, + 0.060253217816352844, + -0.012216916307806969, + 0.0020068222656846046, + 0.06310081481933594, + 0.11251245439052582, + 0.006837606430053711, + 0.03133728355169296, + 0.015237577259540558, + 0.08132436871528625, + -0.007178660482168198, + 0.07070250809192657, + -0.055152423679828644, + 0.015945086255669594, + -0.08277289569377899, + -0.054327234625816345, + -0.08769827336072922, + -0.023721184581518173, + -0.07243412733078003, + -0.1276523917913437, + 0.08239589631557465 + ], + "max_norm_match_direction_error": 0.06125962734222412, + "max_norm_match_relative_error": 1.6587028994763386e-07, + "mean_feedback_forward_cosine": 0.9995171655308116, + "mean_feedback_forward_relative_error": 0.02991201341660185, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.3185271620750427, + "max_absolute_post_projection_soma_slope": 6.10175234783128e-08, + "max_absolute_pre_projection_soma_slope": 0.3185271620750427, + "max_positive_post_projection_soma_slope": 2.269476695460071e-08, + "min_post_projection_soma_slope": -6.10175234783128e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.385685311622679e-08, + "pre_projection_traffic_rms_ratio": 0.01358986675291948 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.013589866752904233, + "raw_negative_gradient_cosine": [ + -0.00474204309284687, + -0.009176000952720642, + -0.0043294597417116165, + -0.0001360562164336443, + -0.0013429553946480155, + -0.005891598295420408, + 0.0028759310953319073, + 0.0023201066069304943, + 0.0017326513770967722, + -0.010575078427791595, + 0.000717283459380269, + -0.03010343573987484, + 0.008269624784588814, + 0.0316404290497303, + 0.006767967715859413, + -0.011222894303500652, + 0.013338077813386917, + 0.020049583166837692, + -0.0010996183846145868, + 0.0243087075650692, + 0.004063806962221861, + 0.03761480748653412, + -0.0053861187770962715, + 0.010938838124275208, + -0.011171959340572357, + 0.038992468267679214, + -0.021670546382665634, + -0.007791978772729635, + -0.013821092434227467, + -0.020795803517103195, + -0.008783942088484764, + 0.024180587381124496, + -0.007901962846517563, + 0.02275543287396431, + -0.012234559282660484, + 0.06025322154164314, + -0.012216916307806969, + 0.0020068208687007427, + 0.06310081481933594, + 0.11251245439052582, + 0.006837606895714998, + 0.03133728355169296, + 0.015237577259540558, + 0.08132436871528625, + -0.007178662810474634, + 0.07070250809192657, + -0.055152423679828644, + 0.015945084393024445, + -0.08277289569377899, + -0.05432722717523575, + -0.08769827336072922, + -0.023721186444163322, + -0.07243412733078003, + -0.1276523917913437, + 0.08239328116178513 + ], + "teaching_negative_gradient_cosine": [ + 0.9995517730712891, + 0.9994828701019287, + 0.9995512962341309, + 0.9991498589515686, + 0.9995520114898682, + 0.9991917610168457, + 0.9995540380477905, + 0.9992571473121643, + 0.9995548725128174, + 0.9994254112243652, + 0.9995452761650085, + 0.9993563294410706, + 0.9995415210723877, + 0.9993796348571777, + 0.9995396733283997, + 0.9994672536849976, + 0.9995462894439697, + 0.9995050430297852, + 0.9995566606521606, + 0.9996000528335571, + 0.9995860457420349, + 0.9994322061538696, + 0.999593198299408, + 0.9994602203369141, + 0.9996090531349182, + 0.9994994401931763, + 0.9996220469474792, + 0.999465823173523, + 0.9996377825737, + 0.9994761943817139, + 0.9996447563171387, + 0.9994639158248901, + 0.9996585249900818, + 0.9994835257530212, + 0.9996689558029175, + 0.9994775652885437, + 0.9996813535690308, + 0.9997172951698303, + 0.9997225999832153, + 0.9996993541717529, + 0.9997093677520752, + 0.9996503591537476, + 0.9997342228889465, + 0.9997210502624512, + 0.9997873306274414, + 0.9998279809951782, + 0.9997203946113586, + 0.9997026920318604, + 0.9997576475143433, + 0.9997735023498535, + 0.9997548460960388, + 0.9996984601020813, + 0.9998241066932678, + 0.9998534321784973, + 0.6879556775093079 + ], + "traffic_instruction_rms_ratio": 863.5765279212795, + "used_negative_gradient_cosine": [ + 0.9995517730712891, + 0.9994828701019287, + 0.9995512962341309, + 0.9991498589515686, + 0.9995520114898682, + 0.9991917610168457, + 0.9995540380477905, + 0.9992571473121643, + 0.9995548725128174, + 0.9994254112243652, + 0.9995452761650085, + 0.9993563294410706, + 0.9995415210723877, + 0.9993796348571777, + 0.9995396733283997, + 0.9994672536849976, + 0.9995462894439697, + 0.9995050430297852, + 0.9995566606521606, + 0.9996000528335571, + 0.9995860457420349, + 0.9994322061538696, + 0.999593198299408, + 0.9994602203369141, + 0.9996090531349182, + 0.9994994401931763, + 0.9996220469474792, + 0.999465823173523, + 0.9996377825737, + 0.9994761943817139, + 0.9996447563171387, + 0.9994639158248901, + 0.9996585249900818, + 0.9994835257530212, + 0.9996689558029175, + 0.9994775652885437, + 0.9996813535690308, + 0.9997172951698303, + 0.9997225999832153, + 0.9996993541717529, + 0.9997093677520752, + 0.9996503591537476, + 0.9997342228889465, + 0.9997210502624512, + 0.9997873306274414, + 0.9998279809951782, + 0.9997203946113586, + 0.9997026920318604, + 0.9997576475143433, + 0.9997735023498535, + 0.9997548460960388, + 0.9996984601020813, + 0.9998241066932678, + 0.9998534321784973, + 0.6879556775093079 + ], + "wall_s": 0.322892427444458 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3656059503555298, + "mean_feedback_forward_cosine": 0.2780039561065761, + "mean_feedback_forward_relative_error": 1.191816060109572, + "min_feedback_forward_cosine": 0.07297273725271225 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0012383521059362043, + "instruction_rms": 0.0012383521056076473, + "raw_apical_rms": 0.062077170118180196, + "teaching_rms": 0.0012383521059362043, + "traffic_rms": 0.061972417824942616 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.997263106242826e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.038032455888949e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01528890456335584, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.4247902322006225 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.349313497543335, + "mean_feedback_forward_cosine": 0.3085195252841169, + "mean_feedback_forward_relative_error": 1.1662700539285487, + "min_feedback_forward_cosine": 0.09686195105314255 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0015625542083360334, + "instruction_rms": 0.0015625542080978654, + "raw_apical_rms": 0.0647349052382139, + "teaching_rms": 0.0015625542083360334, + "traffic_rms": 0.06471562735127863 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.78245655194587e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8353785156196248e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01568088805127374, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.739611852684021 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3408284187316895, + "mean_feedback_forward_cosine": 0.35123352747071873, + "mean_feedback_forward_relative_error": 1.130219289931384, + "min_feedback_forward_cosine": 0.10917021334171295 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0020998698734864697, + "instruction_rms": 0.00209986987322, + "raw_apical_rms": 0.07341691782328486, + "teaching_rms": 0.0020998698734864697, + "traffic_rms": 0.07338650162690992 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.380875644003027e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8044233741425863e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017958872510181766, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4575234595489501 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3346256017684937, + "mean_feedback_forward_cosine": 0.39396569647572255, + "mean_feedback_forward_relative_error": 1.0929852848703212, + "min_feedback_forward_cosine": 0.11670972406864166 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002362502671113251, + "instruction_rms": 0.0023625026705140686, + "raw_apical_rms": 0.08325823013125352, + "teaching_rms": 0.002362502671113251, + "traffic_rms": 0.08322439413299129 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.3774025831025938e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.740517988830108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015586677343912836, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.1788327906036378 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.325390100479126, + "mean_feedback_forward_cosine": 0.43371534862301564, + "mean_feedback_forward_relative_error": 1.0568589454347437, + "min_feedback_forward_cosine": 0.12936991453170776 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002500525164360428, + "instruction_rms": 0.002500525163724199, + "raw_apical_rms": 0.09057189779143957, + "teaching_rms": 0.002500525164360428, + "traffic_rms": 0.0905369907189953 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.751850171824117e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.705191106169133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014525962877544906, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.990774517326355 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3128007650375366, + "mean_feedback_forward_cosine": 0.46883868006142704, + "mean_feedback_forward_relative_error": 1.0236695381728085, + "min_feedback_forward_cosine": 0.1463254690170288 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0024482982780984013, + "instruction_rms": 0.0024482982786286594, + "raw_apical_rms": 0.09678797046191212, + "teaching_rms": 0.0024482982780984013, + "traffic_rms": 0.09675673632942636 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.789049926212101e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6715608148607838e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01364391053927342, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8577928354263306 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2972744703292847, + "mean_feedback_forward_cosine": 0.5014720342376016, + "mean_feedback_forward_relative_error": 0.9913389785723252, + "min_feedback_forward_cosine": 0.16676178574562073 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025669876696848577, + "instruction_rms": 0.002566987669043939, + "raw_apical_rms": 0.10020332202097157, + "teaching_rms": 0.0025669876696848577, + "traffic_rms": 0.10017022184779448 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.267638034771153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6514756405542426e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01372426897854814, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7673397560310363 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.280198097229004, + "mean_feedback_forward_cosine": 0.5318921636451375, + "mean_feedback_forward_relative_error": 0.959966105222702, + "min_feedback_forward_cosine": 0.18302591145038605 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026591966614860343, + "instruction_rms": 0.0026591966620119075, + "raw_apical_rms": 0.10455353882122058, + "teaching_rms": 0.0026591966614860343, + "traffic_rms": 0.10451935582240823 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.183398930190378e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6535967294067978e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013672592862510687, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.7078843341064454 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.266043782234192, + "mean_feedback_forward_cosine": 0.559382903304967, + "mean_feedback_forward_relative_error": 0.9305767151442441, + "min_feedback_forward_cosine": 0.19995196163654327 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002662297389946976, + "instruction_rms": 0.002662297389910144, + "raw_apical_rms": 0.10784836044773859, + "teaching_rms": 0.002662297389946976, + "traffic_rms": 0.10781511058904689 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.459607959641289e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6298795986876625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013918335513955474, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6503286095619202 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2509193420410156, + "mean_feedback_forward_cosine": 0.5853184789419175, + "mean_feedback_forward_relative_error": 0.9019360038367185, + "min_feedback_forward_cosine": 0.21881626546382904 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026984451798573975, + "instruction_rms": 0.0026984451784220187, + "raw_apical_rms": 0.1109552374531632, + "teaching_rms": 0.0026984451798573975, + "traffic_rms": 0.11092206955151283 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.522321337522953e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6258844145573968e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014512654725803377, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.6127370658111573 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2353297472000122, + "mean_feedback_forward_cosine": 0.609443508494984, + "mean_feedback_forward_relative_error": 0.8742235167460009, + "min_feedback_forward_cosine": 0.23856034874916077 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002689612017011485, + "instruction_rms": 0.00268961201754415, + "raw_apical_rms": 0.11372233125296714, + "teaching_rms": 0.002689612017011485, + "traffic_rms": 0.11369013227887856 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.486481479190843e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.620901256783135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01471398521344927, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5740423461914063 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2179542779922485, + "mean_feedback_forward_cosine": 0.6321417673067613, + "mean_feedback_forward_relative_error": 0.8475156426429749, + "min_feedback_forward_cosine": 0.25978827476501465 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027311671063769516, + "instruction_rms": 0.002731167106639558, + "raw_apical_rms": 0.11594257186135726, + "teaching_rms": 0.0027311671063769516, + "traffic_rms": 0.11591006318028584 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.610661694774535e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.62514314116246e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014455018565687839, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5391063870620727 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1979261636734009, + "mean_feedback_forward_cosine": 0.6535548356446352, + "mean_feedback_forward_relative_error": 0.8213582280007276, + "min_feedback_forward_cosine": 0.2839428782463074 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027657178528242322, + "instruction_rms": 0.002765717853341484, + "raw_apical_rms": 0.11822554037526463, + "teaching_rms": 0.0027657178528242322, + "traffic_rms": 0.11819279230943064 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 7.813596880623663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6094369652854546e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014614508352490653, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.5105461913108825 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1780186891555786, + "mean_feedback_forward_cosine": 0.6739871870387685, + "mean_feedback_forward_relative_error": 0.7956407089125026, + "min_feedback_forward_cosine": 0.3079090118408203 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028034998480591833, + "instruction_rms": 0.0028034998492737773, + "raw_apical_rms": 0.12020596693016934, + "teaching_rms": 0.0028034998480591833, + "traffic_rms": 0.12017289592196335 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.659836853690649e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6123015594886733e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015431081585114, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.49121246395111084 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1624243259429932, + "mean_feedback_forward_cosine": 0.6933874021876942, + "mean_feedback_forward_relative_error": 0.7704766788265922, + "min_feedback_forward_cosine": 0.3335566520690918 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027911377137473546, + "instruction_rms": 0.002791137713952103, + "raw_apical_rms": 0.12198513484092295, + "teaching_rms": 0.0027911377137473546, + "traffic_rms": 0.12195282379208204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.941965745634661e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.602053301701936e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015988032490942425, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.4719832165622711 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1462867259979248, + "mean_feedback_forward_cosine": 0.7111540794372558, + "mean_feedback_forward_relative_error": 0.7466255740685896, + "min_feedback_forward_cosine": 0.3583933711051941 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027831336119445234, + "instruction_rms": 0.00278313361201447, + "raw_apical_rms": 0.12361631247874855, + "teaching_rms": 0.0027831336119445234, + "traffic_rms": 0.12358454503316343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.04481246063915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6123693811805665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01525242096805122, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4427167798233032 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1299537420272827, + "mean_feedback_forward_cosine": 0.7280240411108191, + "mean_feedback_forward_relative_error": 0.7232116311788559, + "min_feedback_forward_cosine": 0.38237687945365906 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028365423422506537, + "instruction_rms": 0.002836542340607995, + "raw_apical_rms": 0.12590199217051548, + "teaching_rms": 0.0028365423422506537, + "traffic_rms": 0.12586958803767653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.291042519203984e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6026459912994664e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015464867957198447, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.42650607981681826 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1118658781051636, + "mean_feedback_forward_cosine": 0.7446038468317552, + "mean_feedback_forward_relative_error": 0.6997361838817596, + "min_feedback_forward_cosine": 0.4034211337566376 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002819635216439872, + "instruction_rms": 0.002819635217702689, + "raw_apical_rms": 0.12700602966476648, + "teaching_rms": 0.002819635216439872, + "traffic_rms": 0.12697421958150198 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.838657640173551e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6011112022968033e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015284392847171169, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.411290473651886 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0939488410949707, + "mean_feedback_forward_cosine": 0.760051534392617, + "mean_feedback_forward_relative_error": 0.6772180904041637, + "min_feedback_forward_cosine": 0.4214558005332947 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028309598110479238, + "instruction_rms": 0.0028309598108214136, + "raw_apical_rms": 0.12865474820939268, + "teaching_rms": 0.0028309598110479238, + "traffic_rms": 0.12862305782524822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.623191405376929e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5963102576132327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015575412602281482, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.3980034026527405 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0768272876739502, + "mean_feedback_forward_cosine": 0.7745202443816445, + "mean_feedback_forward_relative_error": 0.6554601989009163, + "min_feedback_forward_cosine": 0.4398435950279236 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028516471527111655, + "instruction_rms": 0.0028516471530947544, + "raw_apical_rms": 0.13114304172775315, + "teaching_rms": 0.0028516471527111655, + "traffic_rms": 0.13111158056093417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.567325065518162e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5852533668031734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015490268180528406, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3804183331108093 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0588077306747437, + "mean_feedback_forward_cosine": 0.7883659655397589, + "mean_feedback_forward_relative_error": 0.6339549872008237, + "min_feedback_forward_cosine": 0.4592060148715973 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028550233654900887, + "instruction_rms": 0.002855023365955546, + "raw_apical_rms": 0.13223170049768876, + "teaching_rms": 0.0028550233654900887, + "traffic_rms": 0.13220037178491406 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.723397660834962e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.586654916052254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015065018997348316, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.36566992959976197 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0408307313919067, + "mean_feedback_forward_cosine": 0.8017013100060549, + "mean_feedback_forward_relative_error": 0.6126981223171407, + "min_feedback_forward_cosine": 0.47839057445526123 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002871351234949905, + "instruction_rms": 0.0028713512359280335, + "raw_apical_rms": 0.13333977765308985, + "teaching_rms": 0.002871351234949905, + "traffic_rms": 0.13330832801217804 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.608424195699627e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5753690378684592e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015327162605551992, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3599096685695648 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0216072797775269, + "mean_feedback_forward_cosine": 0.8140043497085572, + "mean_feedback_forward_relative_error": 0.5924946787682447, + "min_feedback_forward_cosine": 0.4981752634048462 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002868834473953173, + "instruction_rms": 0.0028688344740140826, + "raw_apical_rms": 0.1345024360152318, + "teaching_rms": 0.002868834473953173, + "traffic_rms": 0.13447123155578897 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.339693463061849e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5724156077746804e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015332615251920806, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.34824215928077695 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0027002096176147, + "mean_feedback_forward_cosine": 0.8254355853254145, + "mean_feedback_forward_relative_error": 0.5731189581480893, + "min_feedback_forward_cosine": 0.5162635445594788 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028419032465908916, + "instruction_rms": 0.002841903246743255, + "raw_apical_rms": 0.13575703572374662, + "teaching_rms": 0.0028419032465908916, + "traffic_rms": 0.13572678201609706 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.47092477379374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5739430869928093e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015726947288872844, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3379736206245422 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9825236201286316, + "mean_feedback_forward_cosine": 0.8367859688672152, + "mean_feedback_forward_relative_error": 0.5534052886746147, + "min_feedback_forward_cosine": 0.5360997319221497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002817620692330174, + "instruction_rms": 0.0028176206909344198, + "raw_apical_rms": 0.1368579900648721, + "teaching_rms": 0.002817620692330174, + "traffic_rms": 0.13682842155303404 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.74699779274124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5710004017832103e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01554205884955512, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.3306319559288025 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9619225859642029, + "mean_feedback_forward_cosine": 0.8473270600492304, + "mean_feedback_forward_relative_error": 0.5345866593447599, + "min_feedback_forward_cosine": 0.5557321310043335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028440421257590154, + "instruction_rms": 0.0028440421256435778, + "raw_apical_rms": 0.1380711156486829, + "teaching_rms": 0.0028440421257590154, + "traffic_rms": 0.13804112727234216 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 7.716428740422998e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.572609135623696e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015401260413917114, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.31143450550556184 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9412152171134949, + "mean_feedback_forward_cosine": 0.8572651733051647, + "mean_feedback_forward_relative_error": 0.5162670018998059, + "min_feedback_forward_cosine": 0.574942409992218 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00279067814414045, + "instruction_rms": 0.0027906781417000076, + "raw_apical_rms": 0.13870639797284748, + "teaching_rms": 0.00279067814414045, + "traffic_rms": 0.13867770929227227 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.1187167814341592e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.565223148617519e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015521981619396896, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.3096764917373657 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.918723464012146, + "mean_feedback_forward_cosine": 0.8667979197068648, + "mean_feedback_forward_relative_error": 0.49817959557880054, + "min_feedback_forward_cosine": 0.5950304865837097 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002798082653204898, + "instruction_rms": 0.0027980826532966966, + "raw_apical_rms": 0.13956727491489415, + "teaching_rms": 0.002798082653204898, + "traffic_rms": 0.13953863533790123 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.994531614485823e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.554190116069979e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015236109709625512, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.30386610385894774 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8963318467140198, + "mean_feedback_forward_cosine": 0.8757049127058549, + "mean_feedback_forward_relative_error": 0.4807183177633719, + "min_feedback_forward_cosine": 0.6147742867469788 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002799236645884436, + "instruction_rms": 0.0027992366472855026, + "raw_apical_rms": 0.14078094807955072, + "teaching_rms": 0.002799236645884436, + "traffic_rms": 0.14075251892290597 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 8.548219909698673e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5647263140263365e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015397564473173522, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.29550923011779784 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.874015748500824, + "mean_feedback_forward_cosine": 0.8839899995110252, + "mean_feedback_forward_relative_error": 0.4639662345701998, + "min_feedback_forward_cosine": 0.6334888935089111 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002804685737909153, + "instruction_rms": 0.002804685738499082, + "raw_apical_rms": 0.14221151232407428, + "teaching_rms": 0.002804685737909153, + "traffic_rms": 0.14218327308631767 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.967570946400883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.556512417993898e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015000353483131883, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.28458474427223207 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8500713109970093, + "mean_feedback_forward_cosine": 0.8919064738533714, + "mean_feedback_forward_relative_error": 0.44745709706436504, + "min_feedback_forward_cosine": 0.6534852981567383 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002800285769823482, + "instruction_rms": 0.002800285768673174, + "raw_apical_rms": 0.1434088644149483, + "teaching_rms": 0.002800285769823482, + "traffic_rms": 0.14338099178484853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.120288524400166e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5515314073496795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01509612093560348, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.28211001901626587 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8269674777984619, + "mean_feedback_forward_cosine": 0.8993922114372254, + "mean_feedback_forward_relative_error": 0.4313063205643134, + "min_feedback_forward_cosine": 0.6718124151229858 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002855944463462274, + "instruction_rms": 0.0028559444622426365, + "raw_apical_rms": 0.1439116730699515, + "teaching_rms": 0.002855944463462274, + "traffic_rms": 0.1438826977627302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.415835222625901e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5476991047387974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015297268457968801, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.2752809430599213 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.803813636302948, + "mean_feedback_forward_cosine": 0.9064079057086598, + "mean_feedback_forward_relative_error": 0.4155553829940883, + "min_feedback_forward_cosine": 0.6899509429931641 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027980970268515607, + "instruction_rms": 0.002798097027407999, + "raw_apical_rms": 0.14513959631287332, + "teaching_rms": 0.0027980970268515607, + "traffic_rms": 0.14511196288366107 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0503670466732729e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5410115394510824e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01542476701197429, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.26575205562591553 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7810649871826172, + "mean_feedback_forward_cosine": 0.9129228927872398, + "mean_feedback_forward_relative_error": 0.4004775961691683, + "min_feedback_forward_cosine": 0.7070543766021729 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027667916513980043, + "instruction_rms": 0.0027667916514141915, + "raw_apical_rms": 0.14636993098924692, + "teaching_rms": 0.0027667916513980043, + "traffic_rms": 0.1463432293574295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.262448065148419e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5534027634056982e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01574315852062888, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.26551216046333315 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7573658227920532, + "mean_feedback_forward_cosine": 0.918914317000996, + "mean_feedback_forward_relative_error": 0.38623597811568866, + "min_feedback_forward_cosine": 0.7243421673774719 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027793758233629357, + "instruction_rms": 0.0027793758234007405, + "raw_apical_rms": 0.14706744824421508, + "teaching_rms": 0.0027793758233629357, + "traffic_rms": 0.1470405290119055 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.211240043465295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5396424046847107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015606798978234818, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25689091048240664 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7346653342247009, + "mean_feedback_forward_cosine": 0.9246041319587014, + "mean_feedback_forward_relative_error": 0.372092572667382, + "min_feedback_forward_cosine": 0.7407503128051758 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027499784664082624, + "instruction_rms": 0.0027499784652088534, + "raw_apical_rms": 0.1472130430488538, + "teaching_rms": 0.0027499784664082624, + "traffic_rms": 0.14718676986620913 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.682445778480542e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5375593358201757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015176571577195923, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25194505088329316 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7122654914855957, + "mean_feedback_forward_cosine": 0.9299303098158402, + "mean_feedback_forward_relative_error": 0.35849040909246965, + "min_feedback_forward_cosine": 0.7561821341514587 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002748425787503717, + "instruction_rms": 0.0027484257877978062, + "raw_apical_rms": 0.1484000035842627, + "teaching_rms": 0.002748425787503717, + "traffic_rms": 0.14837394786519092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0385248572220007e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5312229399586802e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014463825279714937, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2506612028312683 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6899703145027161, + "mean_feedback_forward_cosine": 0.9350775090130893, + "mean_feedback_forward_relative_error": 0.3448978062380444, + "min_feedback_forward_cosine": 0.7711887359619141 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027950710030302762, + "instruction_rms": 0.002795071003911944, + "raw_apical_rms": 0.14956033732737076, + "teaching_rms": 0.0027950710030302762, + "traffic_rms": 0.14953354718406292 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.760272779933075e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5331261250516387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014275243696067759, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.25242684220314027 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6686302423477173, + "mean_feedback_forward_cosine": 0.9396645589308306, + "mean_feedback_forward_relative_error": 0.33232562284577977, + "min_feedback_forward_cosine": 0.7849680185317993 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002743590974446751, + "instruction_rms": 0.0027435909752769987, + "raw_apical_rms": 0.1502472012798959, + "teaching_rms": 0.002743590974446751, + "traffic_rms": 0.15022154317834963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0317038601215245e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.521988725490327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014480292337997401, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23705289229393006 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.64716637134552, + "mean_feedback_forward_cosine": 0.9440274401144548, + "mean_feedback_forward_relative_error": 0.319969663430344, + "min_feedback_forward_cosine": 0.7983568906784058 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002753420074918075, + "instruction_rms": 0.0027534200745759855, + "raw_apical_rms": 0.15140253657440755, + "teaching_rms": 0.002753420074918075, + "traffic_rms": 0.15137680800599027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.856902956542399e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5250308844511787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014782177982055547, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23535995511054994 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6272185444831848, + "mean_feedback_forward_cosine": 0.9479562520980835, + "mean_feedback_forward_relative_error": 0.30832505930553783, + "min_feedback_forward_cosine": 0.8104434013366699 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027586520833127494, + "instruction_rms": 0.002758652082026311, + "raw_apical_rms": 0.15205679158888544, + "teaching_rms": 0.0027586520833127494, + "traffic_rms": 0.1520311166437533 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.321422318933401e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5140780827257007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014824770729894528, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22708810456752776 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6066642999649048, + "mean_feedback_forward_cosine": 0.9516334869644859, + "mean_feedback_forward_relative_error": 0.2971851972016421, + "min_feedback_forward_cosine": 0.8226064443588257 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00270368587767172, + "instruction_rms": 0.002703685878245482, + "raw_apical_rms": 0.15260512689355377, + "teaching_rms": 0.00270368587767172, + "traffic_rms": 0.15258049633244022 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.299506421531987e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5165925824567682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014856510001274347, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2268886034297943 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5846050977706909, + "mean_feedback_forward_cosine": 0.9553058591755954, + "mean_feedback_forward_relative_error": 0.28559970625422215, + "min_feedback_forward_cosine": 0.8350316286087036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028172207627640412, + "instruction_rms": 0.0028172207631056824, + "raw_apical_rms": 0.1533917694809535, + "teaching_rms": 0.0028172207627640412, + "traffic_rms": 0.15336523896630078 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.065195717359529e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.513922986618935e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014044546291850594, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22666596289634705 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5634883046150208, + "mean_feedback_forward_cosine": 0.9586459745060314, + "mean_feedback_forward_relative_error": 0.274625505100597, + "min_feedback_forward_cosine": 0.8466119170188904 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027768455455139934, + "instruction_rms": 0.002776845546432808, + "raw_apical_rms": 0.1544028667407089, + "teaching_rms": 0.0027768455455139934, + "traffic_rms": 0.15437715658858833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.297063658346815e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5143296694998058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014141671865212545, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.22244149144172667 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5444114804267883, + "mean_feedback_forward_cosine": 0.9616829536177895, + "mean_feedback_forward_relative_error": 0.2642389884049242, + "min_feedback_forward_cosine": 0.856689453125 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002759754028466141, + "instruction_rms": 0.002759754027834078, + "raw_apical_rms": 0.1553648406338728, + "teaching_rms": 0.002759754028466141, + "traffic_rms": 0.15533962363847936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.780706485187693e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5131244936588388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014700988812350914, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21883933533668518 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5251621007919312, + "mean_feedback_forward_cosine": 0.9645349394191395, + "mean_feedback_forward_relative_error": 0.2541466689922593, + "min_feedback_forward_cosine": 0.8665794134140015 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00279823329800091, + "instruction_rms": 0.002798233298335805, + "raw_apical_rms": 0.1559105639799751, + "teaching_rms": 0.00279823329800091, + "traffic_rms": 0.15588471235802967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.65553798739893e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5018915315216684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014655520919231653, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21692971934318542 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5068506002426147, + "mean_feedback_forward_cosine": 0.9671297506852583, + "mean_feedback_forward_relative_error": 0.24464391551234504, + "min_feedback_forward_cosine": 0.8757303953170776 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002761370938044686, + "instruction_rms": 0.0027613709387155384, + "raw_apical_rms": 0.15661917080913176, + "teaching_rms": 0.002761370938044686, + "traffic_rms": 0.15659412229523495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.148946761821207e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5060615469662893e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014265537423685358, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20904439455032348 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48769253492355347, + "mean_feedback_forward_cosine": 0.9695728464560075, + "mean_feedback_forward_relative_error": 0.23535664142532783, + "min_feedback_forward_cosine": 0.8847986459732056 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027706926316239094, + "instruction_rms": 0.0027706926316897807, + "raw_apical_rms": 0.15763185058879692, + "teaching_rms": 0.0027706926316239094, + "traffic_rms": 0.15760675248286138 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.241346532462558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5103488038343534e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014411582149835166, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.21171879997253418 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4694367051124573, + "mean_feedback_forward_cosine": 0.9718325625766407, + "mean_feedback_forward_relative_error": 0.22650010978633708, + "min_feedback_forward_cosine": 0.8931773900985718 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027425214859031527, + "instruction_rms": 0.002742521484878426, + "raw_apical_rms": 0.15848001441154225, + "teaching_rms": 0.0027425214859031527, + "traffic_rms": 0.15845567246823328 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.1726636017783676e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5056095542910244e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014409455637519409, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20690942641973495 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4519951641559601, + "mean_feedback_forward_cosine": 0.9739293011752042, + "mean_feedback_forward_relative_error": 0.21792619966647842, + "min_feedback_forward_cosine": 0.9008963108062744 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027323529338498225, + "instruction_rms": 0.0027323529340544856, + "raw_apical_rms": 0.15886152125753375, + "teaching_rms": 0.0027323529338498225, + "traffic_rms": 0.15883737040116558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 7.819192404667774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5042070779633913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015131409950644398, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2007670001220703 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4347010552883148, + "mean_feedback_forward_cosine": 0.9759257955984636, + "mean_feedback_forward_relative_error": 0.20939414311539042, + "min_feedback_forward_cosine": 0.9082639217376709 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002800296335199562, + "instruction_rms": 0.0028002963351247668, + "raw_apical_rms": 0.1591612797250569, + "teaching_rms": 0.002800296335199562, + "traffic_rms": 0.15913585408264258 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.851375016798556e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5194261868174983e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014932903840990846, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.2028310550069809 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4180796444416046, + "mean_feedback_forward_cosine": 0.9777542558583346, + "mean_feedback_forward_relative_error": 0.2013039380311966, + "min_feedback_forward_cosine": 0.9149971604347229 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002762755971311803, + "instruction_rms": 0.0027627559718032225, + "raw_apical_rms": 0.15991442728986457, + "teaching_rms": 0.002762755971311803, + "traffic_rms": 0.1598898491344305 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.362687881595775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4959845576227944e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014970237855349773, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.1990632757949829 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40292200446128845, + "mean_feedback_forward_cosine": 0.979403018951416, + "mean_feedback_forward_relative_error": 0.19365910440683365, + "min_feedback_forward_cosine": 0.9209874868392944 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027088232822000237, + "instruction_rms": 0.002708823282228319, + "raw_apical_rms": 0.16119461770656643, + "teaching_rms": 0.0027088232822000237, + "traffic_rms": 0.1611711924001176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.562708421142815e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4954326272514505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014845306560663466, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19544393265724183 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38738200068473816, + "mean_feedback_forward_cosine": 0.9810209491036155, + "mean_feedback_forward_relative_error": 0.18586597415533931, + "min_feedback_forward_cosine": 0.9268696308135986 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027953769088326736, + "instruction_rms": 0.0027953769097796236, + "raw_apical_rms": 0.16208287329107793, + "teaching_rms": 0.0027953769088326736, + "traffic_rms": 0.16205801607526515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0494315461073711e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4951424270734096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013989079206351692, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19609787446498872 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37285012006759644, + "mean_feedback_forward_cosine": 0.9824591062285684, + "mean_feedback_forward_relative_error": 0.17871306172826074, + "min_feedback_forward_cosine": 0.9321953058242798 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027113340226088993, + "instruction_rms": 0.0027113340220307133, + "raw_apical_rms": 0.16335144728010118, + "teaching_rms": 0.0027113340226088993, + "traffic_rms": 0.16332823419035084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.580787470511495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4845757296773175e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014213061413081193, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19339118243217468 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3584863543510437, + "mean_feedback_forward_cosine": 0.983813780004328, + "mean_feedback_forward_relative_error": 0.17169378867203539, + "min_feedback_forward_cosine": 0.9372421503067017 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002794796402549992, + "instruction_rms": 0.002794796402878254, + "raw_apical_rms": 0.16366447011463664, + "teaching_rms": 0.002794796402549992, + "traffic_rms": 0.1636397745953492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.45141966515439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.485367261009298e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01454614872275912, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.19217132482528687 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3457229733467102, + "mean_feedback_forward_cosine": 0.9849926991896196, + "mean_feedback_forward_relative_error": 0.16531535305760123, + "min_feedback_forward_cosine": 0.9416497349739075 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002709675851936232, + "instruction_rms": 0.0027096758522014556, + "raw_apical_rms": 0.16411753845785385, + "teaching_rms": 0.002709675851936232, + "traffic_rms": 0.16409442044698527 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.25869435061577e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.485210677191885e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013679322627510692, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.18285078244686126 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3324398398399353, + "mean_feedback_forward_cosine": 0.9861352920532227, + "mean_feedback_forward_relative_error": 0.15890098667957567, + "min_feedback_forward_cosine": 0.9459847211837769 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002769247779752728, + "instruction_rms": 0.0027692477805658076, + "raw_apical_rms": 0.16498621219585505, + "teaching_rms": 0.002769247779752728, + "traffic_rms": 0.164962176073903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.545276676206413e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4883336961967817e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014030377436603125, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.18484324563026427 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3194241225719452, + "mean_feedback_forward_cosine": 0.9871935801072554, + "mean_feedback_forward_relative_error": 0.15272793200882998, + "min_feedback_forward_cosine": 0.950086236000061 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002771301020854051, + "instruction_rms": 0.0027713010191486175, + "raw_apical_rms": 0.16584593433673436, + "teaching_rms": 0.002771301020854051, + "traffic_rms": 0.165822036554764 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.906699295290309e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4827091109131133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013959305068095439, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18599985842466354 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30728742480278015, + "mean_feedback_forward_cosine": 0.9881513227115978, + "mean_feedback_forward_relative_error": 0.1469086192548275, + "min_feedback_forward_cosine": 0.9537777900695801 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027324350730478625, + "instruction_rms": 0.002732435072453857, + "raw_apical_rms": 0.16627089694018674, + "teaching_rms": 0.0027324350730478625, + "traffic_rms": 0.16624775320576105 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0202614930676646e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.483577913144204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013551371198360728, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.18231251134872437 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2963620722293854, + "mean_feedback_forward_cosine": 0.9890022407878529, + "mean_feedback_forward_relative_error": 0.14155254628170622, + "min_feedback_forward_cosine": 0.9569754600524902 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002649074310180105, + "instruction_rms": 0.0026490743103316796, + "raw_apical_rms": 0.16653584267038904, + "teaching_rms": 0.002649074310180105, + "traffic_rms": 0.16651402756769215 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.703628230297909e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.490160932571076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013738283250032606, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.1762676159477234 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28409332036972046, + "mean_feedback_forward_cosine": 0.9898466478694569, + "mean_feedback_forward_relative_error": 0.1360717303373597, + "min_feedback_forward_cosine": 0.9604353904724121 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026612198871544582, + "instruction_rms": 0.0026612198871531325, + "raw_apical_rms": 0.1669773787694486, + "teaching_rms": 0.0026612198871544582, + "traffic_rms": 0.16695547028302957 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.925126427859141e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4819601401918015e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014204399920962323, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.17534834663391113 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27266091108322144, + "mean_feedback_forward_cosine": 0.99063256112012, + "mean_feedback_forward_relative_error": 0.1307328892702406, + "min_feedback_forward_cosine": 0.9635297060012817 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002726520553956482, + "instruction_rms": 0.002726520554130374, + "raw_apical_rms": 0.16791549876865755, + "teaching_rms": 0.002726520553956482, + "traffic_rms": 0.16789269759356837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.737803914802498e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4861972469913814e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013860654888206024, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.17833706500530244 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2617502808570862, + "mean_feedback_forward_cosine": 0.991372613473372, + "mean_feedback_forward_relative_error": 0.12549196928739548, + "min_feedback_forward_cosine": 0.9663599729537964 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027527037484578777, + "instruction_rms": 0.002752703748279867, + "raw_apical_rms": 0.1682368677981339, + "teaching_rms": 0.0027527037484578777, + "traffic_rms": 0.16821359607107617 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.610852120227719e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787594645407745e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013676186629378301, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17980283549785614 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25183790922164917, + "mean_feedback_forward_cosine": 0.9920222932642156, + "mean_feedback_forward_relative_error": 0.12069282907653939, + "min_feedback_forward_cosine": 0.9688432216644287 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026980213667608137, + "instruction_rms": 0.002698021367352877, + "raw_apical_rms": 0.16910116638988212, + "teaching_rms": 0.0026980213667608137, + "traffic_rms": 0.16907888995223272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.353664071591993e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4927762513104578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013397180396250905, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.16966076889038087 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24140170216560364, + "mean_feedback_forward_cosine": 0.9926310181617737, + "mean_feedback_forward_relative_error": 0.11605323180556297, + "min_feedback_forward_cosine": 0.9713505506515503 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00268975859800863, + "instruction_rms": 0.0026897585972020435, + "raw_apical_rms": 0.16924226494491737, + "teaching_rms": 0.00268975859800863, + "traffic_rms": 0.16922013151353218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 9.400395128977834e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484186332991852e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013206250937574026, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.1738714980792999 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23137395083904266, + "mean_feedback_forward_cosine": 0.9932113582437688, + "mean_feedback_forward_relative_error": 0.11139761940999464, + "min_feedback_forward_cosine": 0.9736660718917847 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027615035597109406, + "instruction_rms": 0.0027615035589753736, + "raw_apical_rms": 0.1696391361692829, + "teaching_rms": 0.0027615035597109406, + "traffic_rms": 0.16961580751596048 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.008461722080028e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4874075596198353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013540547212560579, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.17528437712192535 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22231976687908173, + "mean_feedback_forward_cosine": 0.9937230966307901, + "mean_feedback_forward_relative_error": 0.10713021345436573, + "min_feedback_forward_cosine": 0.9756683707237244 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027198069009940805, + "instruction_rms": 0.002719806901472302, + "raw_apical_rms": 0.1701319062056541, + "teaching_rms": 0.0027198069009940805, + "traffic_rms": 0.17010938667453185 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.012031571524858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4784396601402282e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013908450391089558, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.16779858752965926 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2130720615386963, + "mean_feedback_forward_cosine": 0.9942134954712608, + "mean_feedback_forward_relative_error": 0.10289012596688488, + "min_feedback_forward_cosine": 0.9776421785354614 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002719123901737557, + "instruction_rms": 0.002719123902034706, + "raw_apical_rms": 0.1711123085290241, + "teaching_rms": 0.002719123901737557, + "traffic_rms": 0.17108988808181466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.234762560732634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475695927462311e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013881189615492073, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.16863387867212296 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20451901853084564, + "mean_feedback_forward_cosine": 0.994648505340923, + "mean_feedback_forward_relative_error": 0.09896870546720245, + "min_feedback_forward_cosine": 0.9793894290924072 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027178038617538735, + "instruction_rms": 0.0027178038615526608, + "raw_apical_rms": 0.17167656862859487, + "teaching_rms": 0.0027178038617538735, + "traffic_rms": 0.17165431479285118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.62523563682771e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787696077786304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013782210784177927, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16830873312473296 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1960361897945404, + "mean_feedback_forward_cosine": 0.9950759074904701, + "mean_feedback_forward_relative_error": 0.09497265748002312, + "min_feedback_forward_cosine": 0.981059193611145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027965765732784493, + "instruction_rms": 0.0027965765732825333, + "raw_apical_rms": 0.17181351933746153, + "teaching_rms": 0.0027965765732784493, + "traffic_rms": 0.17178990525302726 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.060342710474288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4909745602320567e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013593266867400998, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.16640550255775452 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18855610489845276, + "mean_feedback_forward_cosine": 0.9954456166787581, + "mean_feedback_forward_relative_error": 0.09135522314093329, + "min_feedback_forward_cosine": 0.9824724197387695 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027281655904317238, + "instruction_rms": 0.002728165590009075, + "raw_apical_rms": 0.17222570224381353, + "teaching_rms": 0.0027281655904317238, + "traffic_rms": 0.17220324485990987 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.835666420736743e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4780435908262375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013637031197511276, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.1663925800704956 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18132393062114716, + "mean_feedback_forward_cosine": 0.9957909085533836, + "mean_feedback_forward_relative_error": 0.08783371133560484, + "min_feedback_forward_cosine": 0.9837785959243774 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027428638790387098, + "instruction_rms": 0.0027428638788873326, + "raw_apical_rms": 0.17282361603639285, + "teaching_rms": 0.0027428638790387098, + "traffic_rms": 0.17280096720411223 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.486171137496058e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474751135245934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01356387755819981, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.1617213433980942 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17441777884960175, + "mean_feedback_forward_cosine": 0.9961136991327459, + "mean_feedback_forward_relative_error": 0.0844164221801541, + "min_feedback_forward_cosine": 0.9849828481674194 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002725872525401691, + "instruction_rms": 0.00272587252533895, + "raw_apical_rms": 0.1733747928635444, + "teaching_rms": 0.002725872525401691, + "traffic_rms": 0.17335269822321894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.193410311856496e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4762007753748884e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01346381957427793, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16320614542007447 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16717448830604553, + "mean_feedback_forward_cosine": 0.9964139970866117, + "mean_feedback_forward_relative_error": 0.08111655596982349, + "min_feedback_forward_cosine": 0.9862015247344971 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027550895293376575, + "instruction_rms": 0.0027550895295905898, + "raw_apical_rms": 0.17407806095253098, + "teaching_rms": 0.0027550895293376575, + "traffic_rms": 0.17405550700181446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.981630917834991e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4767503167194134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013143080538177629, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.16003038954734802 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16055156290531158, + "mean_feedback_forward_cosine": 0.9966846812855114, + "mean_feedback_forward_relative_error": 0.07800691466439855, + "min_feedback_forward_cosine": 0.9872677326202393 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002721852508484442, + "instruction_rms": 0.002721852508214667, + "raw_apical_rms": 0.17447519277667103, + "teaching_rms": 0.002721852508484442, + "traffic_rms": 0.1744530985699871 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.557208820368032e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4709106317474236e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0127308706385453, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.1573057245016098 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1540725976228714, + "mean_feedback_forward_cosine": 0.996940302848816, + "mean_feedback_forward_relative_error": 0.0749533537775278, + "min_feedback_forward_cosine": 0.9882680177688599 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002737049481805662, + "instruction_rms": 0.0027370494821755535, + "raw_apical_rms": 0.17542279218203907, + "teaching_rms": 0.002737049481805662, + "traffic_rms": 0.17540064531710065 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.502558846932516e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4779283795149354e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01281913329776501, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.16388744305610656 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14819373190402985, + "mean_feedback_forward_cosine": 0.9971684510057622, + "mean_feedback_forward_relative_error": 0.0721030096438798, + "min_feedback_forward_cosine": 0.9891375303268433 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027187359129440493, + "instruction_rms": 0.0027187359133989853, + "raw_apical_rms": 0.17567647494748057, + "teaching_rms": 0.0027187359129440493, + "traffic_rms": 0.17565442496905376 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.766786407932159e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.472653181573568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013910260964815483, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15380675556659698 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14218495786190033, + "mean_feedback_forward_cosine": 0.9973890033635227, + "mean_feedback_forward_relative_error": 0.06925406794656407, + "min_feedback_forward_cosine": 0.9899941682815552 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027695661530255608, + "instruction_rms": 0.002769566152455814, + "raw_apical_rms": 0.17628585084616138, + "teaching_rms": 0.0027695661530255608, + "traffic_rms": 0.17626319348051503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.885676384627004e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4804187614589238e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013693297486784349, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.1582255811738968 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13633933663368225, + "mean_feedback_forward_cosine": 0.9975941408764232, + "mean_feedback_forward_relative_error": 0.06649312824010849, + "min_feedback_forward_cosine": 0.9907970428466797 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002816443475617679, + "instruction_rms": 0.0028164434745754324, + "raw_apical_rms": 0.17651873253268258, + "teaching_rms": 0.002816443475617679, + "traffic_rms": 0.17649544806129455 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.836195775074884e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4776062872552448e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013568798378198278, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15971647807121278 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1313091665506363, + "mean_feedback_forward_cosine": 0.9977666052904997, + "mean_feedback_forward_relative_error": 0.06408045044676824, + "min_feedback_forward_cosine": 0.991461992263794 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026407254986483366, + "instruction_rms": 0.0026407254984191324, + "raw_apical_rms": 0.1768421612900572, + "teaching_rms": 0.0026407254986483366, + "traffic_rms": 0.17682161735287463 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.604838086512245e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4750597750868478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013482243886124444, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.14958190220832826 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12596727907657623, + "mean_feedback_forward_cosine": 0.9979402867230502, + "mean_feedback_forward_relative_error": 0.061555661142549734, + "min_feedback_forward_cosine": 0.9921414852142334 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027770749142055247, + "instruction_rms": 0.0027770749138346946, + "raw_apical_rms": 0.17729232448758242, + "teaching_rms": 0.0027770749142055247, + "traffic_rms": 0.17726971160536648 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.676942459511338e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477037823749352e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013216332561294319, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.1503305899810791 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12075788527727127, + "mean_feedback_forward_cosine": 0.9981023419987072, + "mean_feedback_forward_relative_error": 0.05909944095394828, + "min_feedback_forward_cosine": 0.9927743673324585 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027421034410646814, + "instruction_rms": 0.002742103440960747, + "raw_apical_rms": 0.17836122529460033, + "teaching_rms": 0.0027421034410646814, + "traffic_rms": 0.1783391954318167 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.687667391564901e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476841083896905e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014139136984147658, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15631730215907097 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11584215611219406, + "mean_feedback_forward_cosine": 0.9982463218949058, + "mean_feedback_forward_relative_error": 0.056829233281314374, + "min_feedback_forward_cosine": 0.9933488368988037 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002733981069118043, + "instruction_rms": 0.002733981069395037, + "raw_apical_rms": 0.17897381275636734, + "teaching_rms": 0.002733981069118043, + "traffic_rms": 0.1789520942569645 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.952802777959732e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4690582559870694e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013670564581838391, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.15403930703639984 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11129321902990341, + "mean_feedback_forward_cosine": 0.9983807542107322, + "mean_feedback_forward_relative_error": 0.054618410308930006, + "min_feedback_forward_cosine": 0.9938575029373169 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027457106491164396, + "instruction_rms": 0.0027457106498583496, + "raw_apical_rms": 0.17913610256710727, + "teaching_rms": 0.0027457106491164396, + "traffic_rms": 0.17911421019294793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.258651635311253e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.486437144529513e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013436325847240235, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15093926156044007 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10685274749994278, + "mean_feedback_forward_cosine": 0.9985050949183377, + "mean_feedback_forward_relative_error": 0.05248885933648456, + "min_feedback_forward_cosine": 0.9943358898162842 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027675196974467245, + "instruction_rms": 0.0027675196976880046, + "raw_apical_rms": 0.17966310333465813, + "teaching_rms": 0.0027675196974467245, + "traffic_rms": 0.17964092644966032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.142341556549582e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474463248558557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013681089267608467, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.15181832854270935 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10260812193155289, + "mean_feedback_forward_cosine": 0.9986199075525457, + "mean_feedback_forward_relative_error": 0.05044293654235926, + "min_feedback_forward_cosine": 0.9947757124900818 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002714479538790874, + "instruction_rms": 0.0027144795386669306, + "raw_apical_rms": 0.1809547267877564, + "teaching_rms": 0.002714479538790874, + "traffic_rms": 0.18093363378600413 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.150609349082515e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.470303059947909e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013470725174212725, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15021988181114196 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09838725626468658, + "mean_feedback_forward_cosine": 0.9987257534807379, + "mean_feedback_forward_relative_error": 0.04848912277004935, + "min_feedback_forward_cosine": 0.9951949715614319 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026812181871343505, + "instruction_rms": 0.0026812181878185593, + "raw_apical_rms": 0.18103681352514167, + "teaching_rms": 0.0026812181871343505, + "traffic_rms": 0.1810161570403811 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.191571592737091e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476541914775157e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012505334955436063, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14981868672132492 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09436847269535065, + "mean_feedback_forward_cosine": 0.9988214546983892, + "mean_feedback_forward_relative_error": 0.04664448133923791, + "min_feedback_forward_cosine": 0.9955790638923645 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027246578404634034, + "instruction_rms": 0.002724657840126929, + "raw_apical_rms": 0.18158900565248784, + "teaching_rms": 0.0027246578404634034, + "traffic_rms": 0.1815677470554002 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.381829985637523e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4717295155725896e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013407071620656956, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.1476772435569763 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0903562605381012, + "mean_feedback_forward_cosine": 0.9989136370745573, + "mean_feedback_forward_relative_error": 0.04479555529965596, + "min_feedback_forward_cosine": 0.9959458112716675 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002729901653570255, + "instruction_rms": 0.002729901653885861, + "raw_apical_rms": 0.18220239583104256, + "teaching_rms": 0.002729901653570255, + "traffic_rms": 0.18218108333303254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.410380931105465e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477908484232254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013700653902979216, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.15033625076770782 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08651615679264069, + "mean_feedback_forward_cosine": 0.9990009351210161, + "mean_feedback_forward_relative_error": 0.04296609019352631, + "min_feedback_forward_cosine": 0.996281623840332 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028168221669191583, + "instruction_rms": 0.00281682216639033, + "raw_apical_rms": 0.18269745894631845, + "teaching_rms": 0.0028168221669191583, + "traffic_rms": 0.18267486630613974 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.671721308748602e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475657228341853e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013226246337446913, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.15319530076026916 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08301443606615067, + "mean_feedback_forward_cosine": 0.9990757909688083, + "mean_feedback_forward_relative_error": 0.04133343322371895, + "min_feedback_forward_cosine": 0.9965757131576538 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002741218684649971, + "instruction_rms": 0.0027412186840930623, + "raw_apical_rms": 0.18344821290575125, + "teaching_rms": 0.002741218684649971, + "traffic_rms": 0.18342683756432754 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.202172014331154e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4802403303871695e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013582906900337904, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.14606405753135682 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07967378944158554, + "mean_feedback_forward_cosine": 0.9991478063843467, + "mean_feedback_forward_relative_error": 0.03970056924630295, + "min_feedback_forward_cosine": 0.9968456029891968 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027613844665013263, + "instruction_rms": 0.0027613844660805817, + "raw_apical_rms": 0.18359066223267229, + "teaching_rms": 0.0027613844665013263, + "traffic_rms": 0.18356903554737306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.438127624936897e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4873409058114115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013782502611206556, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.152138819065094 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0765121653676033, + "mean_feedback_forward_cosine": 0.9992136185819452, + "mean_feedback_forward_relative_error": 0.038145190697502, + "min_feedback_forward_cosine": 0.9970903396606445 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002739033966105734, + "instruction_rms": 0.002739033966018853, + "raw_apical_rms": 0.18414058605728173, + "teaching_rms": 0.002739033966105734, + "traffic_rms": 0.18411936686464303 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.031909632360112e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.481215366262269e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013539664616705643, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.1473185959672928 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07365334033966064, + "mean_feedback_forward_cosine": 0.999269915710796, + "mean_feedback_forward_relative_error": 0.03676067224957726, + "min_feedback_forward_cosine": 0.9973033666610718 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026663776882964387, + "instruction_rms": 0.0026663776883776974, + "raw_apical_rms": 0.1847611946893068, + "teaching_rms": 0.0026663776882964387, + "traffic_rms": 0.18474113543512805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.161806081512623e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4750452610353226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013702039140369618, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.13559998624801636 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07069694250822067, + "mean_feedback_forward_cosine": 0.9993245666677302, + "mean_feedback_forward_relative_error": 0.03536833387545564, + "min_feedback_forward_cosine": 0.997515082359314 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002737499596678907, + "instruction_rms": 0.002737499596655163, + "raw_apical_rms": 0.1849795838171578, + "teaching_rms": 0.002737499596678907, + "traffic_rms": 0.1849585359227466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.950389774829091e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.486030878707235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014102535988978588, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1400851474881172 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06784866005182266, + "mean_feedback_forward_cosine": 0.9993775053457781, + "mean_feedback_forward_relative_error": 0.033959559368138964, + "min_feedback_forward_cosine": 0.9977103471755981 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027968036200206213, + "instruction_rms": 0.0027968036196581196, + "raw_apical_rms": 0.18526733704289222, + "teaching_rms": 0.0027968036200206213, + "traffic_rms": 0.18524530163454114 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.462853718858241e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478544868037782e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013204779593328695, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.1430394840335846 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06498070061206818, + "mean_feedback_forward_cosine": 0.9994269490242005, + "mean_feedback_forward_relative_error": 0.03259081039577723, + "min_feedback_forward_cosine": 0.9978994727134705 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002783547399337609, + "instruction_rms": 0.0027835473991750623, + "raw_apical_rms": 0.18603764615818696, + "teaching_rms": 0.002783547399337609, + "traffic_rms": 0.18601600897915635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.376798632525606e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4771131466535503e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012784071395670008, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.15067201759815216 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06250336021184921, + "mean_feedback_forward_cosine": 0.9994701027870179, + "mean_feedback_forward_relative_error": 0.03134333484552123, + "min_feedback_forward_cosine": 0.9980563521385193 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002724599039477438, + "instruction_rms": 0.002724599040366459, + "raw_apical_rms": 0.18655293767495848, + "teaching_rms": 0.002724599039477438, + "traffic_rms": 0.18653220014882307 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.885793624178405e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4802724312356196e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01219814796310366, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1414000434613228 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059949520975351334, + "mean_feedback_forward_cosine": 0.999510954726826, + "mean_feedback_forward_relative_error": 0.030118964434686032, + "min_feedback_forward_cosine": 0.9982119202613831 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002737153259694741, + "instruction_rms": 0.0027371532602677097, + "raw_apical_rms": 0.18726668952814665, + "teaching_rms": 0.002737153259694741, + "traffic_rms": 0.18724586210902755 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.164010267819322e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.481039427151832e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013183748792510335, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14272400443077088 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05992874130606651, + "mean_feedback_forward_cosine": 0.9995112830942328, + "mean_feedback_forward_relative_error": 0.030108402991159397, + "min_feedback_forward_cosine": 0.9982131123542786 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001942158526233045, + "instruction_rms": 0.0019421585262273777, + "raw_apical_rms": 0.18742113389850917, + "teaching_rms": 0.001942158526233045, + "traffic_rms": 0.18741040806305445 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.291607317938542e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478928885556829e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013072770639194282, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.0779117331790924 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059918954968452454, + "mean_feedback_forward_cosine": 0.9995114456523548, + "mean_feedback_forward_relative_error": 0.030102703881196, + "min_feedback_forward_cosine": 0.9982137680053711 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016545162463045219, + "instruction_rms": 0.001654516246802673, + "raw_apical_rms": 0.1873739915954006, + "teaching_rms": 0.0016545162463045219, + "traffic_rms": 0.1873659977262605 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.97943249861055e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4789263715738958e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013201778299207382, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05188592261314392 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05991050601005554, + "mean_feedback_forward_cosine": 0.9995115756988525, + "mean_feedback_forward_relative_error": 0.0300976721231233, + "min_feedback_forward_cosine": 0.9982143640518188 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015386455417015568, + "instruction_rms": 0.0015386455426977873, + "raw_apical_rms": 0.1873434472411124, + "teaching_rms": 0.0015386455417015568, + "traffic_rms": 0.18733644340142003 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.404440876574881e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4788611071565946e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012979437441559534, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.0426614681828022 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05990184471011162, + "mean_feedback_forward_cosine": 0.9995117230848832, + "mean_feedback_forward_relative_error": 0.03009267092936418, + "min_feedback_forward_cosine": 0.998214840888977 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015226419026568596, + "instruction_rms": 0.0015226419026997146, + "raw_apical_rms": 0.18729334713932938, + "teaching_rms": 0.0015226419026568596, + "traffic_rms": 0.18728636072278207 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.596431655405468e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4790275359790278e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013196748913822103, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03821386901140213 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05989343672990799, + "mean_feedback_forward_cosine": 0.999511863968589, + "mean_feedback_forward_relative_error": 0.03008786141872406, + "min_feedback_forward_cosine": 0.9982154369354248 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014911648197745945, + "instruction_rms": 0.0014911648194554752, + "raw_apical_rms": 0.18727710966058397, + "teaching_rms": 0.0014911648197745945, + "traffic_rms": 0.18727037468629903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.325321860207623e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4833845815605708e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012971464451058182, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03401476447224617 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059886157512664795, + "mean_feedback_forward_cosine": 0.9995119604197416, + "mean_feedback_forward_relative_error": 0.030083465821702373, + "min_feedback_forward_cosine": 0.9982158541679382 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001407267631694595, + "instruction_rms": 0.0014072676320019287, + "raw_apical_rms": 0.18724369797343227, + "teaching_rms": 0.001407267631694595, + "traffic_rms": 0.18723757700423158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.810741292109014e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480429692707427e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012936318221218403, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.02924771062910557 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059878092259168625, + "mean_feedback_forward_cosine": 0.9995120980522849, + "mean_feedback_forward_relative_error": 0.030078950109468264, + "min_feedback_forward_cosine": 0.9982161521911621 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014091902088174285, + "instruction_rms": 0.0014091902089045907, + "raw_apical_rms": 0.18713996151145967, + "teaching_rms": 0.0014091902088174285, + "traffic_rms": 0.18713378946714446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.551010921564739e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478363271183689e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012920172656626386, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.0280844907951355 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059870265424251556, + "mean_feedback_forward_cosine": 0.9995122129266912, + "mean_feedback_forward_relative_error": 0.030074589310044593, + "min_feedback_forward_cosine": 0.9982166290283203 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013765036207045481, + "instruction_rms": 0.0013765036204422634, + "raw_apical_rms": 0.18710077849262072, + "teaching_rms": 0.0013765036207045481, + "traffic_rms": 0.18709476696712027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.667292744097722e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4782049201410408e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013046690004456846, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.024974282487928868 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05986268073320389, + "mean_feedback_forward_cosine": 0.9995123462243514, + "mean_feedback_forward_relative_error": 0.03007036940279332, + "min_feedback_forward_cosine": 0.9982171654701233 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013269877898053342, + "instruction_rms": 0.0013269877895588928, + "raw_apical_rms": 0.18707458193078763, + "teaching_rms": 0.0013269877898053342, + "traffic_rms": 0.18706893113024098 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.827197461869218e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.486567342851373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012876921724621857, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.023494056099653243 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059854306280612946, + "mean_feedback_forward_cosine": 0.9995124892754988, + "mean_feedback_forward_relative_error": 0.03006591448052363, + "min_feedback_forward_cosine": 0.9982177019119263 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013364903235981624, + "instruction_rms": 0.0013364903236780187, + "raw_apical_rms": 0.18699742168755185, + "teaching_rms": 0.0013364903235981624, + "traffic_rms": 0.18699157680826134 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.297103448740017e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747775629095468e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01275732408278947, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.022603564938306807 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05984736606478691, + "mean_feedback_forward_cosine": 0.9995125857266512, + "mean_feedback_forward_relative_error": 0.03006187825548378, + "min_feedback_forward_cosine": 0.9982181787490845 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012741362988401837, + "instruction_rms": 0.0012741362988686721, + "raw_apical_rms": 0.18695188961192608, + "teaching_rms": 0.0012741362988401837, + "traffic_rms": 0.1869465991556412 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.420381902851659e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4763207550230883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012879078854825995, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02025066905528307 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05984063446521759, + "mean_feedback_forward_cosine": 0.9995127103545449, + "mean_feedback_forward_relative_error": 0.03005808801813559, + "min_feedback_forward_cosine": 0.9982185363769531 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001211851594642384, + "instruction_rms": 0.0012118515946074543, + "raw_apical_rms": 0.1870213174644058, + "teaching_rms": 0.001211851594642384, + "traffic_rms": 0.18701628358515104 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.673247720025756e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4807515270563447e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012868138112846223, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.019001536874175072 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05983375385403633, + "mean_feedback_forward_cosine": 0.9995128295638345, + "mean_feedback_forward_relative_error": 0.030054057787426494, + "min_feedback_forward_cosine": 0.9982188940048218 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012431342331099614, + "instruction_rms": 0.0012431342329435686, + "raw_apical_rms": 0.18701586139105783, + "teaching_rms": 0.0012431342331099614, + "traffic_rms": 0.18701061175803 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.258898276380933e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4856184265565294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012804304711976537, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.018098234566748143 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059826791286468506, + "mean_feedback_forward_cosine": 0.9995129346847534, + "mean_feedback_forward_relative_error": 0.030050288597968492, + "min_feedback_forward_cosine": 0.9982192516326904 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011769220868047725, + "instruction_rms": 0.0011769220868356938, + "raw_apical_rms": 0.18701075603023226, + "teaching_rms": 0.0011769220868047725, + "traffic_rms": 0.18700592761018336 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.684667562874893e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476305081287339e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013279910012647883, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01642048273295164 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059819888323545456, + "mean_feedback_forward_cosine": 0.9995130387219515, + "mean_feedback_forward_relative_error": 0.030046287771653045, + "min_feedback_forward_cosine": 0.9982197880744934 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011878503691832585, + "instruction_rms": 0.0011878503695560122, + "raw_apical_rms": 0.18702650495399994, + "teaching_rms": 0.0011878503691832585, + "traffic_rms": 0.1870214362441155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.991736504844084e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4832989631833948e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012855946959085712, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.016788918063938618 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05981239676475525, + "mean_feedback_forward_cosine": 0.999513176354495, + "mean_feedback_forward_relative_error": 0.030042198283428495, + "min_feedback_forward_cosine": 0.9982202053070068 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011914355144478687, + "instruction_rms": 0.0011914355150809308, + "raw_apical_rms": 0.18697411535031594, + "teaching_rms": 0.0011914355144478687, + "traffic_rms": 0.18696912981872635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.369816756863656e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4764932644005627e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012900762161673884, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.015212382887601853 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059805139899253845, + "mean_feedback_forward_cosine": 0.9995132977312261, + "mean_feedback_forward_relative_error": 0.0300381033427336, + "min_feedback_forward_cosine": 0.9982206225395203 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012213281706388106, + "instruction_rms": 0.001221328170695765, + "raw_apical_rms": 0.1869922069510037, + "teaching_rms": 0.0012213281706388106, + "traffic_rms": 0.18698695680316038 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.74767627831352e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484067996776698e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012955308853082255, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01527994748711586 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05979868397116661, + "mean_feedback_forward_cosine": 0.9995134006847035, + "mean_feedback_forward_relative_error": 0.03003442737019875, + "min_feedback_forward_cosine": 0.9982208609580994 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010872546652513219, + "instruction_rms": 0.0010872546651294226, + "raw_apical_rms": 0.18696344326871037, + "teaching_rms": 0.0010872546652513219, + "traffic_rms": 0.18695904982344846 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.472856111689907e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479270072668026e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012845927343174265, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.012936586709916592 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059791892766952515, + "mean_feedback_forward_cosine": 0.999513526396318, + "mean_feedback_forward_relative_error": 0.03003064472736283, + "min_feedback_forward_cosine": 0.9982212781906128 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011245949005850315, + "instruction_rms": 0.0011245949005166316, + "raw_apical_rms": 0.18698949336273102, + "teaching_rms": 0.0011245949005850315, + "traffic_rms": 0.1869850445402407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.682039893097681e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.473847736468357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01284882631526247, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.01333997043684125 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05978462100028992, + "mean_feedback_forward_cosine": 0.9995136466893283, + "mean_feedback_forward_relative_error": 0.03002660001035441, + "min_feedback_forward_cosine": 0.9982218146324158 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011672219064966247, + "instruction_rms": 0.00116722190617777, + "raw_apical_rms": 0.18695562920044495, + "teaching_rms": 0.0011672219064966247, + "traffic_rms": 0.1869508285452412 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.989620508576991e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.486523422517388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01283147568953709, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.013265175231695176 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05977831035852432, + "mean_feedback_forward_cosine": 0.9995137561451305, + "mean_feedback_forward_relative_error": 0.03002310667017644, + "min_feedback_forward_cosine": 0.9982221126556396 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010335177951959636, + "instruction_rms": 0.0010335177949105903, + "raw_apical_rms": 0.18689662388465386, + "teaching_rms": 0.0010335177951959636, + "traffic_rms": 0.18689261654007722 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.156163033845587e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475174602988262e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013017467388209027, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.011464368002712727 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05977212265133858, + "mean_feedback_forward_cosine": 0.9995138547637246, + "mean_feedback_forward_relative_error": 0.03001955994828181, + "min_feedback_forward_cosine": 0.9982225894927979 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010323254613081173, + "instruction_rms": 0.0010323254608726067, + "raw_apical_rms": 0.18683753988647983, + "teaching_rms": 0.0010323254613081173, + "traffic_rms": 0.1868336075065711 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.242380206662347e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47847848639335e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013198205804345971, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.011206012421697378 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05976475775241852, + "mean_feedback_forward_cosine": 0.9995139728892933, + "mean_feedback_forward_relative_error": 0.03001556901091879, + "min_feedback_forward_cosine": 0.9982229471206665 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010820850332512633, + "instruction_rms": 0.0010820850330885998, + "raw_apical_rms": 0.18686842694037029, + "teaching_rms": 0.0010820850332512633, + "traffic_rms": 0.18686394284838323 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 8.395454642595723e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475929619439481e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013019684343596176, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.011554425864219665 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05975733697414398, + "mean_feedback_forward_cosine": 0.9995141018520702, + "mean_feedback_forward_relative_error": 0.030011543529954825, + "min_feedback_forward_cosine": 0.9982233047485352 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010382241128348978, + "instruction_rms": 0.0010382241131115838, + "raw_apical_rms": 0.18687972231241118, + "teaching_rms": 0.0010382241128348978, + "traffic_rms": 0.1868753730498598 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.734478740279883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4781966537846578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012826877755636871, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.010742876959070563 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059750523418188095, + "mean_feedback_forward_cosine": 0.9995142145590349, + "mean_feedback_forward_relative_error": 0.03000788763165474, + "min_feedback_forward_cosine": 0.9982237815856934 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009770586610846114, + "instruction_rms": 0.0009770586615217218, + "raw_apical_rms": 0.18682898097393796, + "teaching_rms": 0.0009770586610846114, + "traffic_rms": 0.1868250047492078 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.1691675894098807e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4778281326869044e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012914393943515647, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.009893537079840899 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05974414944648743, + "mean_feedback_forward_cosine": 0.9995143175125122, + "mean_feedback_forward_relative_error": 0.030004419013857843, + "min_feedback_forward_cosine": 0.9982240796089172 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009546609549323364, + "instruction_rms": 0.0009546609552980272, + "raw_apical_rms": 0.18676554025584674, + "teaching_rms": 0.0009546609549323364, + "traffic_rms": 0.1867618366481656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.334476492815156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4777879126491513e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012994927788981056, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.009487134465277195 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05973798409104347, + "mean_feedback_forward_cosine": 0.9995144172148271, + "mean_feedback_forward_relative_error": 0.030000815786082637, + "min_feedback_forward_cosine": 0.9982244968414307 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009574786130379354, + "instruction_rms": 0.0009574786133968248, + "raw_apical_rms": 0.18672757408331966, + "teaching_rms": 0.0009574786130379354, + "traffic_rms": 0.18672369534797167 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.777641286565995e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47820517504161e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013229827312661474, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.009010958632156252 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05973127856850624, + "mean_feedback_forward_cosine": 0.9995145320892334, + "mean_feedback_forward_relative_error": 0.02999704374339093, + "min_feedback_forward_cosine": 0.9982248544692993 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009346776505828713, + "instruction_rms": 0.0009346776508154248, + "raw_apical_rms": 0.18674504815725856, + "teaching_rms": 0.0009346776505828713, + "traffic_rms": 0.18674116785449116 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.158221558891455e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479605722879489e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012833709426335346, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.008821906943619251 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05972423031926155, + "mean_feedback_forward_cosine": 0.9995146480473605, + "mean_feedback_forward_relative_error": 0.029993300419300796, + "min_feedback_forward_cosine": 0.9982251524925232 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009514862037800135, + "instruction_rms": 0.0009514862034094702, + "raw_apical_rms": 0.18678238505426353, + "teaching_rms": 0.0009514862037800135, + "traffic_rms": 0.18677846665729889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.698138659861797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47258438696578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012914426389876953, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.008695231651067734 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05971641466021538, + "mean_feedback_forward_cosine": 0.999514779177579, + "mean_feedback_forward_relative_error": 0.029989263508468866, + "min_feedback_forward_cosine": 0.9982256293296814 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010160600266760614, + "instruction_rms": 0.0010160600264414298, + "raw_apical_rms": 0.1867869651857213, + "teaching_rms": 0.0010160600266760614, + "traffic_rms": 0.18678268461079994 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.47478318632966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475585825410848e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01314441359315905, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009065071727130562 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05970918387174606, + "mean_feedback_forward_cosine": 0.999514889717102, + "mean_feedback_forward_relative_error": 0.029985283873975276, + "min_feedback_forward_cosine": 0.9982262253761292 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009661025898037708, + "instruction_rms": 0.0009661025898235843, + "raw_apical_rms": 0.18681788362328494, + "teaching_rms": 0.0009661025898037708, + "traffic_rms": 0.186813822268883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.191830940835644e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4791055802352526e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012773178038589579, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.008534954515248537 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05970335379242897, + "mean_feedback_forward_cosine": 0.9995150013403459, + "mean_feedback_forward_relative_error": 0.029982067729261788, + "min_feedback_forward_cosine": 0.9982265830039978 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008438036926844324, + "instruction_rms": 0.0008438036923690686, + "raw_apical_rms": 0.18686724172096972, + "teaching_rms": 0.0008438036926844324, + "traffic_rms": 0.18686399935115203 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.577431742651243e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4805806657566707e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012885105627298568, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007726747665479779 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05969632416963577, + "mean_feedback_forward_cosine": 0.9995151205496354, + "mean_feedback_forward_relative_error": 0.02997830534692515, + "min_feedback_forward_cosine": 0.9982268810272217 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008754425761856176, + "instruction_rms": 0.0008754425758778264, + "raw_apical_rms": 0.18685698217716118, + "teaching_rms": 0.0008754425761856176, + "traffic_rms": 0.18685359205588317 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.988392525248855e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4769579679196083e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013098200219492477, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.0074592534512281415 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05969060957431793, + "mean_feedback_forward_cosine": 0.999515215917067, + "mean_feedback_forward_relative_error": 0.029974869833412496, + "min_feedback_forward_cosine": 0.9982271790504456 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008447449017975129, + "instruction_rms": 0.0008447449017077916, + "raw_apical_rms": 0.18683611507329428, + "teaching_rms": 0.0008447449017975129, + "traffic_rms": 0.18683295092085908 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.631560177718711e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476099138349961e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013019518790294922, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.007171488909311592 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05968421325087547, + "mean_feedback_forward_cosine": 0.9995153145356612, + "mean_feedback_forward_relative_error": 0.02997163701802492, + "min_feedback_forward_cosine": 0.9982275366783142 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008031027025938207, + "instruction_rms": 0.0008031027028113104, + "raw_apical_rms": 0.1868114497798621, + "teaching_rms": 0.0008031027025938207, + "traffic_rms": 0.1868084887542558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.024171987448426e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4797449531694067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012916726895647419, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.0067251249224878846 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0596780925989151, + "mean_feedback_forward_cosine": 0.9995154239914634, + "mean_feedback_forward_relative_error": 0.029968168518759986, + "min_feedback_forward_cosine": 0.9982279539108276 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008529964937779278, + "instruction_rms": 0.0008529964935085915, + "raw_apical_rms": 0.18679995513472875, + "teaching_rms": 0.0008529964937779278, + "traffic_rms": 0.18679665077269908 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.513643479091115e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47842606064267e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012830811433235358, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.007265768485665321 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05966920778155327, + "mean_feedback_forward_cosine": 0.9995155767960981, + "mean_feedback_forward_relative_error": 0.029963476202366026, + "min_feedback_forward_cosine": 0.9982284307479858 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009935707070761945, + "instruction_rms": 0.000993570707176385, + "raw_apical_rms": 0.1867380515360703, + "teaching_rms": 0.0009935707070761945, + "traffic_rms": 0.18673372745798533 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.923507976987821e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747162185046595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013034490396843025, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.008356949128881097 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059661865234375, + "mean_feedback_forward_cosine": 0.9995156819170171, + "mean_feedback_forward_relative_error": 0.02995954684235833, + "min_feedback_forward_cosine": 0.9982288479804993 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008727971507252572, + "instruction_rms": 0.0008727971507459581, + "raw_apical_rms": 0.18668229633173145, + "teaching_rms": 0.0008727971507252572, + "traffic_rms": 0.18667866415227463 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.846252086414097e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4803124001495526e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012817043182741341, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.006986085975468159 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059654057025909424, + "mean_feedback_forward_cosine": 0.9995158271356063, + "mean_feedback_forward_relative_error": 0.029955337188121947, + "min_feedback_forward_cosine": 0.9982293844223022 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008916215312205233, + "instruction_rms": 0.0008916215313126934, + "raw_apical_rms": 0.18668474478455607, + "teaching_rms": 0.0008916215312205233, + "traffic_rms": 0.18668102833299913 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.296318844078996e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4820548931317565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013047124057827408, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.0073235844442993405 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059646934270858765, + "mean_feedback_forward_cosine": 0.9995159441774542, + "mean_feedback_forward_relative_error": 0.029951450558887288, + "min_feedback_forward_cosine": 0.9982298612594604 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008264584041717986, + "instruction_rms": 0.0008264584042567509, + "raw_apical_rms": 0.18666449406742783, + "teaching_rms": 0.0008264584041717986, + "traffic_rms": 0.18666125231604774 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.427439723845964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4773801558700957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012931396647461181, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.006912893565893173 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059639833867549896, + "mean_feedback_forward_cosine": 0.9995160612193021, + "mean_feedback_forward_relative_error": 0.02994780522686514, + "min_feedback_forward_cosine": 0.9982301592826843 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008025286217986797, + "instruction_rms": 0.0008025286216012244, + "raw_apical_rms": 0.18670021862463038, + "teaching_rms": 0.0008025286217986797, + "traffic_rms": 0.18669716872402395 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0326945698579948e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.48019465706862e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012951156374583238, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.0063300214302539825 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0596332922577858, + "mean_feedback_forward_cosine": 0.9995161815123125, + "mean_feedback_forward_relative_error": 0.029944233875721694, + "min_feedback_forward_cosine": 0.9982306957244873 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007985319130632211, + "instruction_rms": 0.0007985319128746597, + "raw_apical_rms": 0.186704982105077, + "teaching_rms": 0.0007985319130632211, + "traffic_rms": 0.186701868772778 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.169553922243722e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4775734750051242e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013023854800304772, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.006180791571773588 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05962587520480156, + "mean_feedback_forward_cosine": 0.9995163115588102, + "mean_feedback_forward_relative_error": 0.02994006782431494, + "min_feedback_forward_cosine": 0.998231053352356 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008553669860490253, + "instruction_rms": 0.000855366986075815, + "raw_apical_rms": 0.18675268359930192, + "teaching_rms": 0.0008553669860490253, + "traffic_rms": 0.18674920379089202 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.213827840108934e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4742100525759286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012921599960795534, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.006438700502663851 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05961909145116806, + "mean_feedback_forward_cosine": 0.9995164307680997, + "mean_feedback_forward_relative_error": 0.029936198483813892, + "min_feedback_forward_cosine": 0.9982312917709351 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007794813725965413, + "instruction_rms": 0.0007794813724653994, + "raw_apical_rms": 0.18671563883460382, + "teaching_rms": 0.0007794813725965413, + "traffic_rms": 0.1867125467508333 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 8.713207222399433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480179096268245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012793352270195617, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006031738423854112 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05961267650127411, + "mean_feedback_forward_cosine": 0.9995165261355313, + "mean_feedback_forward_relative_error": 0.02993262107060714, + "min_feedback_forward_cosine": 0.998231828212738 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007664058471337904, + "instruction_rms": 0.0007664058470282835, + "raw_apical_rms": 0.1867019661882739, + "teaching_rms": 0.0007664058471337904, + "traffic_rms": 0.18669908622918874 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.544322582793029e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.483720475501362e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01292956864003598, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.005748950992189348 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059606052935123444, + "mean_feedback_forward_cosine": 0.999516656182029, + "mean_feedback_forward_relative_error": 0.02992896953259002, + "min_feedback_forward_cosine": 0.9982321262359619 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007540146366630595, + "instruction_rms": 0.0007540146363055531, + "raw_apical_rms": 0.18671151420617393, + "teaching_rms": 0.0007540146366630595, + "traffic_rms": 0.18670857409183306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.464748025791778e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4766567216134175e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012997984364375865, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.005779394161738455 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05959717556834221, + "mean_feedback_forward_cosine": 0.9995167808099227, + "mean_feedback_forward_relative_error": 0.029924737094816835, + "min_feedback_forward_cosine": 0.9982324838638306 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007508879656496231, + "instruction_rms": 0.0007508879658370791, + "raw_apical_rms": 0.18672524164647283, + "teaching_rms": 0.0007508879656496231, + "traffic_rms": 0.18672238315170148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.713833293290008e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4756739745275323e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012784698319785816, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.005757287951242179 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05959020555019379, + "mean_feedback_forward_cosine": 0.9995169097726996, + "mean_feedback_forward_relative_error": 0.02992104718630964, + "min_feedback_forward_cosine": 0.9982329607009888 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007416679675730267, + "instruction_rms": 0.0007416679679445114, + "raw_apical_rms": 0.18665447920841813, + "teaching_rms": 0.0007416679675730267, + "traffic_rms": 0.1866517554820684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.1132273414204974e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4754667044465384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012835192054512574, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.0055073183459788564 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059584714472293854, + "mean_feedback_forward_cosine": 0.9995169975540855, + "mean_feedback_forward_relative_error": 0.029917928991331295, + "min_feedback_forward_cosine": 0.9982333779335022 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007042286171929342, + "instruction_rms": 0.0007042286170597567, + "raw_apical_rms": 0.18664652561908457, + "teaching_rms": 0.0007042286171929342, + "traffic_rms": 0.18664413531554655 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.102095788193765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478297975262729e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013095309411453907, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.005166453071124851 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957819148898125, + "mean_feedback_forward_cosine": 0.9995171167633751, + "mean_feedback_forward_relative_error": 0.02991415344348008, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007263999673380577, + "instruction_rms": 0.0007263999672897956, + "raw_apical_rms": 0.18665944440941087, + "teaching_rms": 0.0007263999673380577, + "traffic_rms": 0.1866566009121289 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.110971805308509e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787928518474366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012837200778514517, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005089225533213467 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957811698317528, + "mean_feedback_forward_cosine": 0.9995171254331415, + "mean_feedback_forward_relative_error": 0.029914089724082837, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006989284992739937, + "instruction_rms": 0.0006989284993360486, + "raw_apical_rms": 0.18664669591787328, + "teaching_rms": 0.0006989284992739937, + "traffic_rms": 0.1866441315551133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.742617214034908e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4790241581807717e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012806894408613864, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005087548696175217 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0595780648291111, + "mean_feedback_forward_cosine": 0.9995171167633751, + "mean_feedback_forward_relative_error": 0.029914041464640335, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006207069902482965, + "instruction_rms": 0.0006207069904778508, + "raw_apical_rms": 0.18664896936985398, + "teaching_rms": 0.0006207069902482965, + "traffic_rms": 0.18664688348841363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 8.849252708387212e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4785809190451758e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012759993664397357, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004274944938607514 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957799404859543, + "mean_feedback_forward_cosine": 0.9995171221819791, + "mean_feedback_forward_relative_error": 0.029913986660540105, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007062200358760151, + "instruction_rms": 0.0007062200359924701, + "raw_apical_rms": 0.18665144051940874, + "teaching_rms": 0.0007062200358760151, + "traffic_rms": 0.1866487272315597 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.908508337095554e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4781292260919833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012765790708509512, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0051097940035536886 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957792326807976, + "mean_feedback_forward_cosine": 0.999517128684304, + "mean_feedback_forward_relative_error": 0.02991392993452874, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006819489927441199, + "instruction_rms": 0.0006819489927718962, + "raw_apical_rms": 0.1866589728178419, + "teaching_rms": 0.0006819489927441199, + "traffic_rms": 0.18665628607406326 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.881925429524927e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747021712676862e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012847949953898541, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004796492007207126 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957786738872528, + "mean_feedback_forward_cosine": 0.9995171308517456, + "mean_feedback_forward_relative_error": 0.02991388225927949, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006392848930633222, + "instruction_rms": 0.0006392848930196327, + "raw_apical_rms": 0.18666418053623685, + "teaching_rms": 0.0006392848930633222, + "traffic_rms": 0.1866618890704653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.299836753787531e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4806397545765177e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012866563626659664, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004376249497383833 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0595778226852417, + "mean_feedback_forward_cosine": 0.9995171254331415, + "mean_feedback_forward_relative_error": 0.02991383436389945, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005940520451153719, + "instruction_rms": 0.0005940520453499666, + "raw_apical_rms": 0.1866558727031781, + "teaching_rms": 0.0005940520451153719, + "traffic_rms": 0.18665384832908294 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.667196027545288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476787288596777e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012817021557280346, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.003956728442832827 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577763080596924, + "mean_feedback_forward_cosine": 0.999517128684304, + "mean_feedback_forward_relative_error": 0.02991378584199331, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006079692369889687, + "instruction_rms": 0.0006079692368506197, + "raw_apical_rms": 0.1866573862281726, + "teaching_rms": 0.0006079692369889687, + "traffic_rms": 0.18665503796565674 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.043279192155751e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747311702899342e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012984927274426546, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004120183801501989 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577710926532745, + "mean_feedback_forward_cosine": 0.9995171330191872, + "mean_feedback_forward_relative_error": 0.029913743229752238, + "min_feedback_forward_cosine": 0.9982336759567261 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000561001494353049, + "instruction_rms": 0.0005610014945096465, + "raw_apical_rms": 0.18665922469500534, + "teaching_rms": 0.000561001494353049, + "traffic_rms": 0.18665758228490972 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.884477699031777e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4775050765567242e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012760971228347037, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0035482280809804798 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957766994833946, + "mean_feedback_forward_cosine": 0.9995171319354664, + "mean_feedback_forward_relative_error": 0.0299137003635141, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005486807378581706, + "instruction_rms": 0.0005486807378516426, + "raw_apical_rms": 0.18665699034052097, + "teaching_rms": 0.0005486807378581706, + "traffic_rms": 0.18665524537012923 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.348493369363496e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478105915108401e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013011781256530252, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.0037513064473122357 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957763269543648, + "mean_feedback_forward_cosine": 0.9995171265168623, + "mean_feedback_forward_relative_error": 0.029913658293133433, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005311206470752708, + "instruction_rms": 0.0005311206471777004, + "raw_apical_rms": 0.18665823243288274, + "teaching_rms": 0.0005311206470752708, + "traffic_rms": 0.1866565836462628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.734904189897861e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4797725964129925e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012783898524403089, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.0037402107955515386 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577569365501404, + "mean_feedback_forward_cosine": 0.9995171232656999, + "mean_feedback_forward_relative_error": 0.02991361182860353, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006192587370477685, + "instruction_rms": 0.0006192587372561701, + "raw_apical_rms": 0.18666022737558965, + "teaching_rms": 0.0006192587370477685, + "traffic_rms": 0.18665791163582418 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.618795822532775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4753387631875118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012841801735557361, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004223174134641886 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957752838730812, + "mean_feedback_forward_cosine": 0.999517146023837, + "mean_feedback_forward_relative_error": 0.029913569055497646, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005418691716150356, + "instruction_rms": 0.0005418691718005687, + "raw_apical_rms": 0.18665334218792606, + "teaching_rms": 0.0005418691716150356, + "traffic_rms": 0.18665158740824273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.374361388201578e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.473903288588127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012873883466402142, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.0035883484309911728 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577491134405136, + "mean_feedback_forward_cosine": 0.9995171210982583, + "mean_feedback_forward_relative_error": 0.029913524876941332, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005269440986083987, + "instruction_rms": 0.0005269440985704833, + "raw_apical_rms": 0.18664855258332727, + "teaching_rms": 0.0005269440986083987, + "traffic_rms": 0.18664696014678983 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.112205224757417e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4759976990524157e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012846993109639817, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0036596060974150896 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957745015621185, + "mean_feedback_forward_cosine": 0.9995171341029081, + "mean_feedback_forward_relative_error": 0.029913484914736315, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00054186018645757, + "instruction_rms": 0.0005418601860191863, + "raw_apical_rms": 0.1866405979912566, + "teaching_rms": 0.00054186018645757, + "traffic_rms": 0.1866388892250604 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.145666813812568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4791799051981314e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012603416204608473, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.003821141163185239 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577398002147675, + "mean_feedback_forward_cosine": 0.9995171210982583, + "mean_feedback_forward_relative_error": 0.029913441743701696, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005452266166730762, + "instruction_rms": 0.000545226616683032, + "raw_apical_rms": 0.18664138808590405, + "teaching_rms": 0.0005452266166730762, + "traffic_rms": 0.1866395575935221 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.621206942256322e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4806005224282078e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012895695003580798, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.003640412979386747 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577349573373795, + "mean_feedback_forward_cosine": 0.999517128684304, + "mean_feedback_forward_relative_error": 0.029913395262238653, + "min_feedback_forward_cosine": 0.9982336759567261 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005915571073394919, + "instruction_rms": 0.0005915571072099585, + "raw_apical_rms": 0.18664666001724223, + "teaching_rms": 0.0005915571073394919, + "traffic_rms": 0.1866445670609501 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.03632068166371e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4793646532554472e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012808791494214477, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.0038668042615428565 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957731232047081, + "mean_feedback_forward_cosine": 0.9995171297680248, + "mean_feedback_forward_relative_error": 0.029913348958573557, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005591776984511721, + "instruction_rms": 0.0005591776981465066, + "raw_apical_rms": 0.18664749065959121, + "teaching_rms": 0.0005591776984511721, + "traffic_rms": 0.18664570914458695 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.815835800784043e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4750950418910274e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012973032973654864, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.0036486537681892513 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957726389169693, + "mean_feedback_forward_cosine": 0.9995171384377913, + "mean_feedback_forward_relative_error": 0.029913308776237747, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005043270512319984, + "instruction_rms": 0.0005043270512382743, + "raw_apical_rms": 0.18663938486460932, + "teaching_rms": 0.0005043270512319984, + "traffic_rms": 0.18663791643239808 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.859103590135419e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4788078616644682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012902646947275394, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.00334836719032377 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957722291350365, + "mean_feedback_forward_cosine": 0.9995171471075578, + "mean_feedback_forward_relative_error": 0.029913269694555888, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005064346710496158, + "instruction_rms": 0.0005064346711717509, + "raw_apical_rms": 0.18663635001236725, + "teaching_rms": 0.0005064346710496158, + "traffic_rms": 0.1866347927415454 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.710053123195394e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47691296540597e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012686280792321026, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0033029225072264672 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957719683647156, + "mean_feedback_forward_cosine": 0.9995171373540704, + "mean_feedback_forward_relative_error": 0.02991323248398575, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000479637529325789, + "instruction_rms": 0.00047963752899314827, + "raw_apical_rms": 0.18663327622993256, + "teaching_rms": 0.000479637529325789, + "traffic_rms": 0.1866318703891748 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0323599752837254e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787497523303574e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012920421547514879, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0031628264583274724 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957716330885887, + "mean_feedback_forward_cosine": 0.9995171395215121, + "mean_feedback_forward_relative_error": 0.029913193427703596, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004757740128272622, + "instruction_rms": 0.00047577401319643574, + "raw_apical_rms": 0.18663403344230053, + "teaching_rms": 0.0004757740128272622, + "traffic_rms": 0.18663267663285907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.355986918535564e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4795457673305356e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012915480081032276, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0032090540381520985 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577129781246185, + "mean_feedback_forward_cosine": 0.999517146023837, + "mean_feedback_forward_relative_error": 0.02991315379569476, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005125808800569914, + "instruction_rms": 0.0005125808798465811, + "raw_apical_rms": 0.18662802418328658, + "teaching_rms": 0.0005125808800569914, + "traffic_rms": 0.18662640119151994 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.141886726458324e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478740091223978e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013032117989121645, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003341278999596834 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059577081352472305, + "mean_feedback_forward_cosine": 0.9995171438563953, + "mean_feedback_forward_relative_error": 0.02991311000660062, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005533441213541519, + "instruction_rms": 0.0005533441214684674, + "raw_apical_rms": 0.18663434927615838, + "teaching_rms": 0.0005533441213541519, + "traffic_rms": 0.186632462252632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.861794332735371e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4774759593591502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012830458537427567, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0037860340731963515 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957702919840813, + "mean_feedback_forward_cosine": 0.9995171406052329, + "mean_feedback_forward_relative_error": 0.029913066073574805, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005469834207018864, + "instruction_rms": 0.0005469834206710233, + "raw_apical_rms": 0.18664213345236574, + "teaching_rms": 0.0005469834207018864, + "traffic_rms": 0.18664039917398684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.471091834077015e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4802466997955682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01282248651170203, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0036384447392076255 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957697704434395, + "mean_feedback_forward_cosine": 0.9995171427726746, + "mean_feedback_forward_relative_error": 0.029913023038005287, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005445844431018095, + "instruction_rms": 0.0005445844431124797, + "raw_apical_rms": 0.18664081313273514, + "teaching_rms": 0.0005445844431018095, + "traffic_rms": 0.18663896853117368 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.421262265301266e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4724891485291135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012868412107575007, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0035656695897318424 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957691743969917, + "mean_feedback_forward_cosine": 0.9995171525261619, + "mean_feedback_forward_relative_error": 0.029912977530197664, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005653015603015497, + "instruction_rms": 0.0005653015600237457, + "raw_apical_rms": 0.18663648643542105, + "teaching_rms": 0.0005653015603015497, + "traffic_rms": 0.1866345467321993 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.040241539220006e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4828347367383505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012889968649969762, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0038163919495418667 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0595768578350544, + "mean_feedback_forward_cosine": 0.9995171438563953, + "mean_feedback_forward_relative_error": 0.029912931311198255, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005489472686096788, + "instruction_rms": 0.0005489472684613421, + "raw_apical_rms": 0.18663049588081557, + "teaching_rms": 0.0005489472686096788, + "traffic_rms": 0.18662858974425778 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.184047490227385e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4769373853317913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012791655765685871, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0036822275079227986 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957682803273201, + "mean_feedback_forward_cosine": 0.9995171579447659, + "mean_feedback_forward_relative_error": 0.029912890984930777, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047704517577872876, + "instruction_rms": 0.0004770451753518119, + "raw_apical_rms": 0.18663196047101704, + "teaching_rms": 0.00047704517577872876, + "traffic_rms": 0.18663051786757903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271918773651, + "maximum_absolute_post_projection_soma_slope": 9.044511983802295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.483580601854625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012872338638024136, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0030486267526447773 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059576790779829025, + "mean_feedback_forward_cosine": 0.999517146023837, + "mean_feedback_forward_relative_error": 0.029912852140312845, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004916926312980989, + "instruction_rms": 0.0004916926307836782, + "raw_apical_rms": 0.18664377083122818, + "teaching_rms": 0.0004916926312980989, + "traffic_rms": 0.18664231961139657 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.265830959113373e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480360351416137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012879682106809603, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0032361116764321922 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957676097750664, + "mean_feedback_forward_cosine": 0.9995171579447659, + "mean_feedback_forward_relative_error": 0.02991280805488879, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005169557290670346, + "instruction_rms": 0.0005169557291345863, + "raw_apical_rms": 0.18664430771649565, + "teaching_rms": 0.0005169557290670346, + "traffic_rms": 0.18664245738929391 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.824790853623199e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478123996651909e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013131541915244235, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0034099118908494713 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957673117518425, + "mean_feedback_forward_cosine": 0.9995171471075578, + "mean_feedback_forward_relative_error": 0.02991276964206587, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004809978813720436, + "instruction_rms": 0.0004809978814631061, + "raw_apical_rms": 0.1866350212683364, + "teaching_rms": 0.0004809978813720436, + "traffic_rms": 0.18663370955945524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.22165935940211e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4778154438063627e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01287879613473309, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0030512185246869923 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059576671570539474, + "mean_feedback_forward_cosine": 0.9995171666145325, + "mean_feedback_forward_relative_error": 0.029912731212309817, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005257758968280134, + "instruction_rms": 0.0005257758967415225, + "raw_apical_rms": 0.18663674269964067, + "teaching_rms": 0.0005257758968280134, + "traffic_rms": 0.1866349901506463 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.27520807433757e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4776303943188244e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012734691925956746, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.003455720066949725 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957663059234619, + "mean_feedback_forward_cosine": 0.9995171481912787, + "mean_feedback_forward_relative_error": 0.029912689980119467, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004981248276489748, + "instruction_rms": 0.0004981248278038391, + "raw_apical_rms": 0.186632755578277, + "teaching_rms": 0.0004981248276489748, + "traffic_rms": 0.18663114211666887 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 9.91904727243309e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480798850718967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012970175365574247, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.003394184242375195 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957659333944321, + "mean_feedback_forward_cosine": 0.99951716336337, + "mean_feedback_forward_relative_error": 0.029912649950181894, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004960422399883992, + "instruction_rms": 0.0004960422399533717, + "raw_apical_rms": 0.18663802116408837, + "teaching_rms": 0.0004960422399883992, + "traffic_rms": 0.18663650201563378 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.959346248753718e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4783344385231284e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012945955778359426, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.003346009315811098 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957654491066933, + "mean_feedback_forward_cosine": 0.999517151442441, + "mean_feedback_forward_relative_error": 0.029912611190229654, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000476197966572041, + "instruction_rms": 0.00047619796655630427, + "raw_apical_rms": 0.1866444932829352, + "teaching_rms": 0.000476197966572041, + "traffic_rms": 0.18664317152787918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.151284447421858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4770518072379014e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012758164968853621, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0032784904893767087 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059576503932476044, + "mean_feedback_forward_cosine": 0.9995171557773244, + "mean_feedback_forward_relative_error": 0.029912570931694724, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004893118856280414, + "instruction_rms": 0.0004893118855300508, + "raw_apical_rms": 0.18664030640370077, + "teaching_rms": 0.0004893118856280414, + "traffic_rms": 0.1866388296615656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.31037204104723e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477576738231598e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012944032869712761, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.00309908518075943 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957647040486336, + "mean_feedback_forward_cosine": 0.9995171601122076, + "mean_feedback_forward_relative_error": 0.029912527929991484, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004702014470020697, + "instruction_rms": 0.00047020144702556615, + "raw_apical_rms": 0.18664710708804347, + "teaching_rms": 0.0004702014470020697, + "traffic_rms": 0.18664574576514556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.792723266675239e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4804154977319748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013042132416354723, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0030700711361132563 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957642197608948, + "mean_feedback_forward_cosine": 0.9995171427726746, + "mean_feedback_forward_relative_error": 0.029912484606558626, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005106957195743687, + "instruction_rms": 0.0005106957194416652, + "raw_apical_rms": 0.18664366227902096, + "teaching_rms": 0.0005106957195743687, + "traffic_rms": 0.18664203880877497 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 1.0748531309445752e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4767191460470957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01290948249769636, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003250807805247605 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957638844847679, + "mean_feedback_forward_cosine": 0.9995171492749995, + "mean_feedback_forward_relative_error": 0.02991244544021108, + "min_feedback_forward_cosine": 0.9982337355613708 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004959199538155213, + "instruction_rms": 0.0004959199538960415, + "raw_apical_rms": 0.18664201711974426, + "teaching_rms": 0.0004959199538155213, + "traffic_rms": 0.18664042968792968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.335487595128143e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4809355058893583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012803780023543383, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003114388564787805 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957634747028351, + "mean_feedback_forward_cosine": 0.9995171492749995, + "mean_feedback_forward_relative_error": 0.02991240465674888, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004921742954982841, + "instruction_rms": 0.0004921742952473915, + "raw_apical_rms": 0.18664724874328475, + "teaching_rms": 0.0004921742954982841, + "traffic_rms": 0.18664571851109546 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 6.77418299233068e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475773008163202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012919296369654188, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.003053068472594023 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059576310217380524, + "mean_feedback_forward_cosine": 0.999517151442441, + "mean_feedback_forward_relative_error": 0.029912363449958238, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048334032593967343, + "instruction_rms": 0.00048334032596021435, + "raw_apical_rms": 0.18665046370645902, + "teaching_rms": 0.00048334032593967343, + "traffic_rms": 0.1866488532616208 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.61868656443221e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479201813981741e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012872271055091658, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0031814973421767354 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957627668976784, + "mean_feedback_forward_cosine": 0.999517168781974, + "mean_feedback_forward_relative_error": 0.02991232926195318, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004676042283203498, + "instruction_rms": 0.0004676042282208105, + "raw_apical_rms": 0.18664621764315356, + "teaching_rms": 0.0004676042283203498, + "traffic_rms": 0.1866447581930081 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.51766799314646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4757499398162268e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012928854062600473, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.002995151945091784 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957621708512306, + "mean_feedback_forward_cosine": 0.999517146023837, + "mean_feedback_forward_relative_error": 0.02991228382187811, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005338150018967615, + "instruction_rms": 0.000533815001992186, + "raw_apical_rms": 0.18665041220911016, + "teaching_rms": 0.0005338150018967615, + "traffic_rms": 0.18664859941901282 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.104821264647398e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4794814320075422e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012866536640122108, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0035769257015921176 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957619473338127, + "mean_feedback_forward_cosine": 0.9995171503587202, + "mean_feedback_forward_relative_error": 0.029912244393066927, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004778134905261213, + "instruction_rms": 0.0004778134905853502, + "raw_apical_rms": 0.1866401666274628, + "teaching_rms": 0.0004778134905261213, + "traffic_rms": 0.18663888882489296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.118527634020211e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4752765880491253e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012873923038486307, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.002965674960911274 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957615748047829, + "mean_feedback_forward_cosine": 0.99951716336337, + "mean_feedback_forward_relative_error": 0.029912206301973623, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004795112649822119, + "instruction_rms": 0.0004795112648570268, + "raw_apical_rms": 0.18664798509075933, + "teaching_rms": 0.0004795112649822119, + "traffic_rms": 0.1866464829699653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.177995797746917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474273393090714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012907226697625722, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0029603301522880794 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0595761239528656, + "mean_feedback_forward_cosine": 0.9995171471075578, + "mean_feedback_forward_relative_error": 0.029912164866585624, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048471119131891717, + "instruction_rms": 0.0004847111912744987, + "raw_apical_rms": 0.18664612704297034, + "teaching_rms": 0.00048471119131891717, + "traffic_rms": 0.18664457282716895 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.759996461549235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4848247804668447e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012945925474393704, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.003223515403866768 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957607552409172, + "mean_feedback_forward_cosine": 0.9995171676982533, + "mean_feedback_forward_relative_error": 0.029912125759504058, + "min_feedback_forward_cosine": 0.9982338547706604 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00046530950113513644, + "instruction_rms": 0.0004653095011022904, + "raw_apical_rms": 0.18665192434996628, + "teaching_rms": 0.00046530950113513644, + "traffic_rms": 0.186650655865962 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.077129993149356e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476141842211579e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012850973888043151, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.002941495413389057 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957602337002754, + "mean_feedback_forward_cosine": 0.9995171698656949, + "mean_feedback_forward_relative_error": 0.029912084849043327, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005266028045938286, + "instruction_rms": 0.0005266028045694331, + "raw_apical_rms": 0.18665276586797971, + "teaching_rms": 0.0005266028045938286, + "traffic_rms": 0.1866509706718818 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.776154913903156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477309332139587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012880499576332546, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0032209800898469984 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059575989842414856, + "mean_feedback_forward_cosine": 0.9995171666145325, + "mean_feedback_forward_relative_error": 0.02991204886612567, + "min_feedback_forward_cosine": 0.9982339143753052 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00043926572547460925, + "instruction_rms": 0.0004392657253135081, + "raw_apical_rms": 0.1866517709466827, + "teaching_rms": 0.00043926572547460925, + "traffic_rms": 0.18665052580726155 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 8.785022487245442e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475849300108685e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012866491229723899, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.002796236093044281 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05957596004009247, + "mean_feedback_forward_cosine": 0.9995171655308116, + "mean_feedback_forward_relative_error": 0.02991201341660185, + "min_feedback_forward_cosine": 0.9982337951660156 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004426267905250608, + "instruction_rms": 0.00044262679066082625, + "raw_apical_rms": 0.186655493469265, + "teaching_rms": 0.0004426267905250608, + "traffic_rms": 0.18665431801409485 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3185271620750427, + "maximum_absolute_post_projection_soma_slope": 7.816931457682585e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4755178700752272e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012914272432439625, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0028824622425809503 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9285, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.3843463958740234, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3324800000, + "peak_memory_reserved_bytes": 5553258496, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 9.786987931192925e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 9.541088985542956e-08, + "min_residual_soma_slope": -9.786987931192925e-08, + "mse": 1.575032135943921e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.15922135791212e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.15922135791212e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 11, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.025099039077759, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.06901049613952637, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 26394.462034463882, + "train_wall_s": 26388.074353456497 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.04923790320754051, + 0.035815004259347916, + 0.02320684678852558, + 0.020803283900022507, + 0.015287086367607117, + 0.016048796474933624, + 0.010450431145727634, + 0.013235553167760372, + 0.008099833503365517, + 0.010244094766676426, + 0.006573145743459463, + 0.0090594831854105, + 0.005592650733888149, + 0.00767307635396719, + 0.0046932571567595005, + 0.006742183584719896, + 0.004148817155510187, + 0.006013964302837849, + 0.0036442363634705544, + 0.008519847877323627, + 0.005758453160524368, + 0.007350224535912275, + 0.004980198107659817, + 0.006826599128544331, + 0.004265956114977598, + 0.006016553845256567, + 0.0037007362116128206, + 0.0052390024065971375, + 0.003234435338526964, + 0.004589166026562452, + 0.0029108552262187004, + 0.004201128147542477, + 0.002640962367877364, + 0.003908427897840738, + 0.0024022667203098536, + 0.003455269616097212, + 0.002206935314461589, + 0.005219635553658009, + 0.0035466987174004316, + 0.004828799050301313, + 0.003267403692007065, + 0.00448852451518178, + 0.0030291208531707525, + 0.004144572652876377, + 0.002798451343551278, + 0.003775671124458313, + 0.002612984273582697, + 0.0035160058178007603, + 0.002493839245289564, + 0.0033689457923173904, + 0.0023912033066153526, + 0.0029892241582274437, + 0.002305891364812851, + 0.002823576331138611, + 0.0023601697757840157 + ], + "realized_traffic_instruction_rms_ratio": [ + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 3.999999761581421, + 4.000000476837158, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.2609078586101532, + 0.19511745870113373, + 0.07969685643911362, + 0.10733510553836823, + 0.04281584173440933, + 0.08453045785427094, + 0.02423582598567009, + 0.06916831433773041, + 0.016409888863563538, + 0.05496395751833916, + 0.011917604133486748, + 0.04795212671160698, + 0.009447183459997177, + 0.04183388501405716, + 0.007368385326117277, + 0.03587426245212555, + 0.006118131335824728, + 0.03148999437689781, + 0.0050909025594592094, + 0.045080699026584625, + 0.010493498295545578, + 0.03868812695145607, + 0.008390847593545914, + 0.03666605427861214, + 0.006721850950270891, + 0.03273526206612587, + 0.0054259453900158405, + 0.027816765010356903, + 0.004501056857407093, + 0.024026326835155487, + 0.0038577693048864603, + 0.021978143602609634, + 0.003324762685224414, + 0.020618751645088196, + 0.0029013685416430235, + 0.0182217787951231, + 0.0025801616720855236, + 0.02806156873703003, + 0.005548752378672361, + 0.02528063766658306, + 0.004763890523463488, + 0.023494817316532135, + 0.004181653261184692, + 0.022056877613067627, + 0.0036565035115927458, + 0.019951466470956802, + 0.0032284297049045563, + 0.018626803532242775, + 0.0029618756379932165, + 0.018041668459773064, + 0.002743529388681054, + 0.01574643887579441, + 0.0025562939699739218, + 0.014796272851526737, + 0.002505142241716385 + ], + "unscaled_traffic_rms": [ + 0.7548703551292419, + 0.7342244982719421, + 1.164755940437317, + 0.7752648591995239, + 1.4281710386276245, + 0.7594326138496399, + 1.7247905731201172, + 0.765411376953125, + 1.9743785858154297, + 0.7455136179924011, + 2.2061970233917236, + 0.7557106614112854, + 2.3679652214050293, + 0.7336709499359131, + 2.5477805137634277, + 0.7517571449279785, + 2.7124733924865723, + 0.7639206647872925, + 2.8633322715759277, + 0.7559641599655151, + 2.1950557231903076, + 0.7599462866783142, + 2.3741097450256348, + 0.7447323799133301, + 2.538560390472412, + 0.7351770997047424, + 2.728178024291992, + 0.7533589601516724, + 2.8743786811828613, + 0.7640228867530823, + 3.018174409866333, + 0.7646011114120483, + 3.1773242950439453, + 0.7582278251647949, + 3.311908483505249, + 0.7584922909736633, + 3.4213907718658447, + 0.7440261840820312, + 2.5567541122436523, + 0.7640312314033508, + 2.7434749603271484, + 0.7641727328300476, + 2.897534132003784, + 0.7516154646873474, + 3.0613412857055664, + 0.7569711208343506, + 3.2374677658081055, + 0.7550422549247742, + 3.3679189682006836, + 0.7469255328178406, + 3.4863169193267822, + 0.7593397498130798, + 3.6081788539886475, + 0.7633209824562073, + 3.7685201168060303 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250441282772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 127796290519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 5010596513857536 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json new file mode 100644 index 0000000..968f855 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json @@ -0,0 +1,7293 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 12, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 12, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5012, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9993400242593553, + "feedback_forward_cosine": [ + 0.9996849298477173, + 0.9996539354324341, + 0.999669075012207, + 0.99958336353302, + 0.9995406866073608, + 0.9995044469833374, + 0.9991855621337891, + 0.9991582632064819, + 0.999467670917511, + 0.9994152784347534, + 0.9996272325515747, + 0.9995869994163513, + 0.9996865391731262, + 0.9996215105056763, + 0.9997698068618774, + 0.9996647834777832, + 0.9996969699859619, + 0.9994884729385376, + 0.9996986389160156, + 0.9997678995132446, + 0.9996581077575684, + 0.9996104836463928, + 0.9996354579925537, + 0.9995818138122559, + 0.9996703863143921, + 0.9995665550231934, + 0.9996948838233948, + 0.9995852112770081, + 0.9996534585952759, + 0.9995052814483643, + 0.9996408820152283, + 0.9994964599609375, + 0.9996259808540344, + 0.9994370937347412, + 0.999616265296936, + 0.9994439482688904, + 0.999738335609436, + 0.9998199939727783, + 0.9997847080230713, + 0.999758780002594, + 0.9997938871383667, + 0.9997113943099976, + 0.9997418522834778, + 0.9996110796928406, + 0.9997273683547974, + 0.9995449185371399, + 0.999617338180542, + 0.9993396997451782, + 0.9991607666015625, + 0.9987689256668091, + 0.9988960027694702, + 0.9984460473060608, + 0.99857497215271, + 0.9979240894317627, + 0.9999796152114868 + ], + "feedback_forward_norm_ratio": [ + 1.00027334690094, + 0.9991751313209534, + 0.9988250136375427, + 1.0006673336029053, + 1.000723958015442, + 1.0003741979599, + 0.9992495775222778, + 1.0013952255249023, + 0.9989585280418396, + 0.9998454451560974, + 0.999127209186554, + 1.0013729333877563, + 0.9994022250175476, + 1.0009852647781372, + 0.998449981212616, + 1.001898169517517, + 0.9997077584266663, + 1.0009105205535889, + 0.9987272620201111, + 1.0004408359527588, + 0.9991752505302429, + 1.0011051893234253, + 0.99944007396698, + 1.001184105873108, + 0.998991072177887, + 1.001473307609558, + 0.9990416169166565, + 1.0016357898712158, + 0.999473512172699, + 1.0016686916351318, + 0.9987295269966125, + 1.001453161239624, + 0.9989418983459473, + 1.001962423324585, + 0.9988632202148438, + 1.0023912191390991, + 0.9995113611221313, + 1.0003210306167603, + 0.9997403621673584, + 1.0008677244186401, + 0.9998303651809692, + 1.0011351108551025, + 0.9994261860847473, + 1.0012418031692505, + 0.9994954466819763, + 1.0017199516296387, + 0.9992617964744568, + 1.0020018815994263, + 0.9983162879943848, + 1.0028022527694702, + 0.9975609183311462, + 1.0023900270462036, + 0.9977555274963379, + 1.0036203861236572, + 1.0019617080688477 + ], + "feedback_forward_relative_error": [ + 0.025110743939876556, + 0.026312218979001045, + 0.02573779784142971, + 0.028882427141070366, + 0.030329221859574318, + 0.03149121627211571, + 0.0403495728969574, + 0.041082851588726044, + 0.03263014927506447, + 0.03419363126158714, + 0.02730858325958252, + 0.028792714700102806, + 0.025037001818418503, + 0.027544578537344933, + 0.021494587883353233, + 0.025987165048718452, + 0.02461758442223072, + 0.032012443989515305, + 0.024566933512687683, + 0.021558070555329323, + 0.026150986552238464, + 0.027946995571255684, + 0.02699824795126915, + 0.028958939015865326, + 0.025685135275125504, + 0.029500126838684082, + 0.024707775563001633, + 0.028874926269054413, + 0.026322131976485252, + 0.031524140387773514, + 0.026816997677087784, + 0.03178944066166878, + 0.027355700731277466, + 0.03364518657326698, + 0.02770969830453396, + 0.03347347676753998, + 0.02287558652460575, + 0.018977820873260498, + 0.02074912190437317, + 0.021991359069943428, + 0.020304592326283455, + 0.0240655355155468, + 0.022720960900187492, + 0.02793707139790058, + 0.02335325814783573, + 0.030243588611483574, + 0.027662605047225952, + 0.03643273562192917, + 0.04096972197294235, + 0.049767762422561646, + 0.04699651151895523, + 0.05586620792746544, + 0.05337286368012428, + 0.06465364992618561, + 0.006678103934973478 + ], + "innovation_negative_gradient_cosine": [ + 0.9993928670883179, + 0.9992727041244507, + 0.9994022846221924, + 0.9992779493331909, + 0.9994056224822998, + 0.9992627501487732, + 0.9994044303894043, + 0.9990179538726807, + 0.9994091987609863, + 0.9991894960403442, + 0.9994136095046997, + 0.9993071556091309, + 0.9994266033172607, + 0.9993363618850708, + 0.9994335770606995, + 0.9993957281112671, + 0.9994494915008545, + 0.9993226528167725, + 0.9994568824768066, + 0.9994272589683533, + 0.9994800686836243, + 0.999329686164856, + 0.9994857311248779, + 0.9993296265602112, + 0.9994916915893555, + 0.9993273019790649, + 0.9994919300079346, + 0.9993712306022644, + 0.9995056390762329, + 0.9993817806243896, + 0.9995151162147522, + 0.9993694424629211, + 0.9995219707489014, + 0.9993258714675903, + 0.999531626701355, + 0.9993797540664673, + 0.9995588064193726, + 0.9995734691619873, + 0.9995944499969482, + 0.9995929598808289, + 0.999625563621521, + 0.9995740652084351, + 0.9996570348739624, + 0.9996242523193359, + 0.9996660947799683, + 0.9996587038040161, + 0.9996583461761475, + 0.9996837377548218, + 0.9996843338012695, + 0.9996918439865112, + 0.9996574521064758, + 0.9996740221977234, + 0.9997631311416626, + 0.9998087882995605, + 0.7778909206390381 + ], + "instruction_negative_gradient_cosine": [ + 0.9993928670883179, + 0.9992727041244507, + 0.9994021654129028, + 0.9992778897285461, + 0.9994056224822998, + 0.9992627501487732, + 0.9994044303894043, + 0.9990179538726807, + 0.9994090795516968, + 0.9991894960403442, + 0.9994136095046997, + 0.9993071556091309, + 0.9994266033172607, + 0.9993363618850708, + 0.9994335174560547, + 0.9993957281112671, + 0.9994494915008545, + 0.9993226528167725, + 0.9994568824768066, + 0.9994272589683533, + 0.9994800090789795, + 0.999329686164856, + 0.9994857311248779, + 0.9993296265602112, + 0.9994916915893555, + 0.9993271827697754, + 0.9994919300079346, + 0.9993711709976196, + 0.9995056390762329, + 0.9993817806243896, + 0.9995151162147522, + 0.9993694424629211, + 0.9995219707489014, + 0.9993258714675903, + 0.999531626701355, + 0.9993797540664673, + 0.9995587468147278, + 0.9995734095573425, + 0.9995944499969482, + 0.9995930194854736, + 0.9996255040168762, + 0.9995740056037903, + 0.9996570348739624, + 0.9996242523193359, + 0.9996660947799683, + 0.9996587038040161, + 0.9996583461761475, + 0.9996837377548218, + 0.9996843338012695, + 0.9996917843818665, + 0.9996574521064758, + 0.9996740221977234, + 0.9997631311416626, + 0.9998087882995605, + 0.78672856092453 + ], + "matched_negative_gradient_cosine": [ + 0.01258007436990738, + 0.011209037154912949, + 0.014262584038078785, + 0.005753850564360619, + 0.01989125832915306, + 0.023908644914627075, + 0.03258507326245308, + 0.06666544079780579, + 0.012788617052137852, + 0.029419491067528725, + 0.0251312218606472, + 0.011835029348731041, + 0.020365310832858086, + 0.03372950851917267, + 0.029591450467705727, + 0.03314148262143135, + 0.04314713925123215, + 0.042947057634592056, + 0.04670741781592369, + 0.08639397472143173, + 0.04796481132507324, + 0.04514092206954956, + 0.0559760257601738, + 0.11523230373859406, + 0.0637044608592987, + 0.0892191231250763, + 0.04783106595277786, + 0.07903582602739334, + 0.05096733570098877, + 0.11316852271556854, + 0.04805481433868408, + 0.14113305509090424, + 0.04574444890022278, + 0.19634845852851868, + 0.018344972282648087, + 0.1734047830104828, + -0.00048715248703956604, + 0.0348823145031929, + 0.0001741647720336914, + 0.16805273294448853, + -0.010281551629304886, + 0.14894333481788635, + 0.02791683003306389, + 0.27264603972435, + 0.042339541018009186, + 0.19911828637123108, + 0.03964167460799217, + 0.1310538649559021, + 0.052645936608314514, + 0.08929893374443054, + 0.06234383583068848, + 0.09579779207706451, + 0.06370040774345398, + 0.06218864768743515, + 0.11700852960348129 + ], + "max_norm_match_direction_error": 1.7881393432617188e-07, + "max_norm_match_relative_error": 1.8084102748616715e-07, + "mean_feedback_forward_cosine": 0.9995060747320001, + "mean_feedback_forward_relative_error": 0.029965790149501778, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.3032892942428589, + "max_absolute_post_projection_soma_slope": 3.358382372198321e-08, + "max_absolute_pre_projection_soma_slope": 0.3032892942428589, + "max_positive_post_projection_soma_slope": 2.8097604243271235e-08, + "min_post_projection_soma_slope": -3.358382372198321e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.450058861831175e-08, + "pre_projection_traffic_rms_ratio": 0.015686577031590324 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.015686577031573362, + "raw_negative_gradient_cosine": [ + 0.01258007436990738, + 0.011209037154912949, + 0.01426258310675621, + 0.005753850564360619, + 0.01989126019179821, + 0.023908644914627075, + 0.03258507326245308, + 0.06666544079780579, + 0.012788617052137852, + 0.029419491067528725, + 0.02513122372329235, + 0.01183503121137619, + 0.020365310832858086, + 0.03372950851917267, + 0.029591452330350876, + 0.03314148262143135, + 0.04314713925123215, + 0.04294705390930176, + 0.04670742154121399, + 0.08639398217201233, + 0.04796481132507324, + 0.04514091834425926, + 0.055976029485464096, + 0.11523230373859406, + 0.0637044608592987, + 0.08921913057565689, + 0.04783106595277786, + 0.07903582602739334, + 0.05096733570098877, + 0.11316852271556854, + 0.04805481433868408, + 0.14113305509090424, + 0.04574444890022278, + 0.19634845852851868, + 0.018344976007938385, + 0.1734047830104828, + -0.0004871548153460026, + 0.0348823145031929, + 0.00017416104674339294, + 0.16805273294448853, + -0.01028155442327261, + 0.14894333481788635, + 0.027916831895709038, + 0.27264603972435, + 0.042339541018009186, + 0.19911827147006989, + 0.039641670882701874, + 0.1310538649559021, + 0.052645936608314514, + 0.08929893374443054, + 0.06234383583068848, + 0.09579779207706451, + 0.06370040774345398, + 0.06218864768743515, + 0.11700852960348129 + ], + "teaching_negative_gradient_cosine": [ + 0.9993928670883179, + 0.9992727041244507, + 0.9994022846221924, + 0.9992779493331909, + 0.9994056224822998, + 0.9992627501487732, + 0.9994044303894043, + 0.9990179538726807, + 0.9994091987609863, + 0.9991894960403442, + 0.9994136095046997, + 0.9993071556091309, + 0.9994266033172607, + 0.9993363618850708, + 0.9994335770606995, + 0.9993957281112671, + 0.9994494915008545, + 0.9993226528167725, + 0.9994568824768066, + 0.9994272589683533, + 0.9994800686836243, + 0.999329686164856, + 0.9994857311248779, + 0.9993296265602112, + 0.9994916915893555, + 0.9993273019790649, + 0.9994919300079346, + 0.9993712306022644, + 0.9995056390762329, + 0.9993817806243896, + 0.9995151162147522, + 0.9993694424629211, + 0.9995219707489014, + 0.9993258714675903, + 0.999531626701355, + 0.9993797540664673, + 0.9995588064193726, + 0.9995734691619873, + 0.9995944499969482, + 0.9995929598808289, + 0.999625563621521, + 0.9995740652084351, + 0.9996570348739624, + 0.9996242523193359, + 0.9996660947799683, + 0.9996587038040161, + 0.9996583461761475, + 0.9996837377548218, + 0.9996843338012695, + 0.9996918439865112, + 0.9996574521064758, + 0.9996740221977234, + 0.9997631311416626, + 0.9998087882995605, + 0.7778909206390381 + ], + "traffic_instruction_rms_ratio": 69.78429475863933, + "used_negative_gradient_cosine": [ + 0.9993928670883179, + 0.9992727041244507, + 0.9994022846221924, + 0.9992779493331909, + 0.9994056224822998, + 0.9992627501487732, + 0.9994044303894043, + 0.9990179538726807, + 0.9994091987609863, + 0.9991894960403442, + 0.9994136095046997, + 0.9993071556091309, + 0.9994266033172607, + 0.9993363618850708, + 0.9994335770606995, + 0.9993957281112671, + 0.9994494915008545, + 0.9993226528167725, + 0.9994568824768066, + 0.9994272589683533, + 0.9994800686836243, + 0.999329686164856, + 0.9994857311248779, + 0.9993296265602112, + 0.9994916915893555, + 0.9993273019790649, + 0.9994919300079346, + 0.9993712306022644, + 0.9995056390762329, + 0.9993817806243896, + 0.9995151162147522, + 0.9993694424629211, + 0.9995219707489014, + 0.9993258714675903, + 0.999531626701355, + 0.9993797540664673, + 0.9995588064193726, + 0.9995734691619873, + 0.9995944499969482, + 0.9995929598808289, + 0.999625563621521, + 0.9995740652084351, + 0.9996570348739624, + 0.9996242523193359, + 0.9996660947799683, + 0.9996587038040161, + 0.9996583461761475, + 0.9996837377548218, + 0.9996843338012695, + 0.9996918439865112, + 0.9996574521064758, + 0.9996740221977234, + 0.9997631311416626, + 0.9998087882995605, + 0.7778909206390381 + ], + "wall_s": 0.3470437526702881 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.345116376876831, + "mean_feedback_forward_cosine": 0.350229903513735, + "mean_feedback_forward_relative_error": 1.1264865322546525, + "min_feedback_forward_cosine": 0.10241095721721649 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.001619350113461001, + "instruction_rms": 0.0016193501132363646, + "raw_apical_rms": 0.07822498465239641, + "teaching_rms": 0.001619350113461001, + "traffic_rms": 0.0780933375962263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3439932899927953e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.030163009286038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014015765537670068, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.3700356150054933 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3402464389801025, + "mean_feedback_forward_cosine": 0.3699487951668826, + "mean_feedback_forward_relative_error": 1.1099011486226862, + "min_feedback_forward_cosine": 0.10818107426166534 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0013438463815941039, + "instruction_rms": 0.001343846381590453, + "raw_apical_rms": 0.0879618686767622, + "teaching_rms": 0.0013438463815941039, + "traffic_rms": 0.08795138626991492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.4986434848651697e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8095191880555645e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014977222345471936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.6639973961639405 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.333116888999939, + "mean_feedback_forward_cosine": 0.4049170349131931, + "mean_feedback_forward_relative_error": 1.0799586458639665, + "min_feedback_forward_cosine": 0.11794281750917435 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0019955159152813826, + "instruction_rms": 0.001995515915862228, + "raw_apical_rms": 0.09899625871126506, + "teaching_rms": 0.0019955159152813826, + "traffic_rms": 0.0989758205310678 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.372140872035743e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.780017113597399e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014874718473815919, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4130293825912477 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3236387968063354, + "mean_feedback_forward_cosine": 0.4455680020830848, + "mean_feedback_forward_relative_error": 1.0433958741751583, + "min_feedback_forward_cosine": 0.13184186816215515 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002393122705073572, + "instruction_rms": 0.0023931227055234787, + "raw_apical_rms": 0.11142030672977384, + "teaching_rms": 0.002393122705073572, + "traffic_rms": 0.11139431630090205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2837897145345778e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7404264829296693e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014630101665655171, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.1560584063720702 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3117445707321167, + "mean_feedback_forward_cosine": 0.48197933354161004, + "mean_feedback_forward_relative_error": 1.008802363547412, + "min_feedback_forward_cosine": 0.14702005684375763 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025510534997813944, + "instruction_rms": 0.002551053499623267, + "raw_apical_rms": 0.12185380512200401, + "teaching_rms": 0.0025510534997813944, + "traffic_rms": 0.12182684807116438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2910211921735026e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6992431153639914e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014134442802580112, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 0.9538153356552124 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3000320196151733, + "mean_feedback_forward_cosine": 0.5142867042259737, + "mean_feedback_forward_relative_error": 0.976585992899808, + "min_feedback_forward_cosine": 0.1623951494693756 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026827340845482607, + "instruction_rms": 0.002682734084591011, + "raw_apical_rms": 0.12955621136754142, + "teaching_rms": 0.0026827340845482607, + "traffic_rms": 0.12952820527480227 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.185546452120434e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6695849325819913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014324208747756351, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8234868240737915 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2874479293823242, + "mean_feedback_forward_cosine": 0.5428382358767769, + "mean_feedback_forward_relative_error": 0.9472316254268993, + "min_feedback_forward_cosine": 0.17822204530239105 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00265320213801056, + "instruction_rms": 0.002653202136959813, + "raw_apical_rms": 0.13483498981907466, + "teaching_rms": 0.00265320213801056, + "traffic_rms": 0.13480860808254905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.2264801796391112e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.659273587704981e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01478570958967619, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7281737655258179 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2742054462432861, + "mean_feedback_forward_cosine": 0.5696408558975566, + "mean_feedback_forward_relative_error": 0.9185948366468603, + "min_feedback_forward_cosine": 0.1957789659500122 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002640756773487682, + "instruction_rms": 0.002640756774457201, + "raw_apical_rms": 0.13905199472353788, + "teaching_rms": 0.002640756773487682, + "traffic_rms": 0.13902669344969967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4947872273296525e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6513175653943744e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01428331577343606, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6628663708877564 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2604798078536987, + "mean_feedback_forward_cosine": 0.5941899982365695, + "mean_feedback_forward_relative_error": 0.891240857406096, + "min_feedback_forward_cosine": 0.2134874165058136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002648417372320954, + "instruction_rms": 0.002648417372028608, + "raw_apical_rms": 0.1425505626211793, + "teaching_rms": 0.002648417372320954, + "traffic_rms": 0.14252566645966236 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8292143977305386e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6481228537427718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014241256105048665, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6140893089866638 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.245855689048767, + "mean_feedback_forward_cosine": 0.6169147122990001, + "mean_feedback_forward_relative_error": 0.8650280052965338, + "min_feedback_forward_cosine": 0.23124197125434875 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026620247851548867, + "instruction_rms": 0.0026620247861943026, + "raw_apical_rms": 0.14557844963083333, + "teaching_rms": 0.0026620247851548867, + "traffic_rms": 0.14555382157775307 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 2.2829946999536332e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6356783581056634e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014394338670337653, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5630594208526611 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2304452657699585, + "mean_feedback_forward_cosine": 0.6381275783885609, + "mean_feedback_forward_relative_error": 0.8399280954491009, + "min_feedback_forward_cosine": 0.2505224347114563 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026323133397672636, + "instruction_rms": 0.002632313338684203, + "raw_apical_rms": 0.14847824705823684, + "teaching_rms": 0.0026323133397672636, + "traffic_rms": 0.14845463218061458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 2.054305667797962e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.622871280500468e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014472494174313334, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5326842555809022 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2172038555145264, + "mean_feedback_forward_cosine": 0.6579306580803611, + "mean_feedback_forward_relative_error": 0.8153891200369054, + "min_feedback_forward_cosine": 0.26771971583366394 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002654071485010499, + "instruction_rms": 0.002654071484907767, + "raw_apical_rms": 0.15034974817227792, + "teaching_rms": 0.002654071485010499, + "traffic_rms": 0.1503260603221622 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.710032222490554e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6345083195984834e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014525893307746287, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5030954770469666 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.205235242843628, + "mean_feedback_forward_cosine": 0.6766121230342171, + "mean_feedback_forward_relative_error": 0.7916638645258817, + "min_feedback_forward_cosine": 0.2862786650657654 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026344707352182464, + "instruction_rms": 0.002634470734875863, + "raw_apical_rms": 0.15271385835072118, + "teaching_rms": 0.0026344707352182464, + "traffic_rms": 0.15269083550051263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.8988072270076373e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.629680186559267e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01436875910092409, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.4757079619026184 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1927727460861206, + "mean_feedback_forward_cosine": 0.6937667727470398, + "mean_feedback_forward_relative_error": 0.7692238236015493, + "min_feedback_forward_cosine": 0.3035103380680084 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026010628361919207, + "instruction_rms": 0.0026010628363993976, + "raw_apical_rms": 0.15486260457574044, + "teaching_rms": 0.0026010628361919207, + "traffic_rms": 0.15484048946350348 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.9605045054049697e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6251122080735277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014669651440180934, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4505086038017273 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1788723468780518, + "mean_feedback_forward_cosine": 0.7107458965344863, + "mean_feedback_forward_relative_error": 0.7464412792162461, + "min_feedback_forward_cosine": 0.3232823610305786 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026137198638953234, + "instruction_rms": 0.0026137198627354417, + "raw_apical_rms": 0.1568845278927749, + "teaching_rms": 0.0026137198638953234, + "traffic_rms": 0.15686242414354754 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7944692842775112e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619787700408085e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0150431062441356, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.43340378182411193 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1645333766937256, + "mean_feedback_forward_cosine": 0.7267557469281283, + "mean_feedback_forward_relative_error": 0.724306046962738, + "min_feedback_forward_cosine": 0.3442862331867218 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002635920754411942, + "instruction_rms": 0.0026359207545455375, + "raw_apical_rms": 0.15805994391209077, + "teaching_rms": 0.002635920754411942, + "traffic_rms": 0.1580376314866191 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5350865112395695e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.615456236104047e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014802625572070814, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.41914446306228637 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1491841077804565, + "mean_feedback_forward_cosine": 0.7420737402005629, + "mean_feedback_forward_relative_error": 0.7026614451950247, + "min_feedback_forward_cosine": 0.3645966351032257 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025979459990690475, + "instruction_rms": 0.0025979460000215477, + "raw_apical_rms": 0.15910876727478795, + "teaching_rms": 0.0025979459990690475, + "traffic_rms": 0.15908719721940232 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.440065346969277e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6111924221687466e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014429671225327524, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.4033851655101776 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1335734128952026, + "mean_feedback_forward_cosine": 0.7568492721427571, + "mean_feedback_forward_relative_error": 0.681088940392841, + "min_feedback_forward_cosine": 0.38252386450767517 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026453039621508424, + "instruction_rms": 0.0026453039629758717, + "raw_apical_rms": 0.16027157372562492, + "teaching_rms": 0.0026453039621508424, + "traffic_rms": 0.16024939379138922 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.7309837119228177e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6095408043039333e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014534682627667064, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3801485876083374 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1178427934646606, + "mean_feedback_forward_cosine": 0.7709077726710927, + "mean_feedback_forward_relative_error": 0.66011607457291, + "min_feedback_forward_cosine": 0.39994385838508606 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002638740849668689, + "instruction_rms": 0.0026387408488585014, + "raw_apical_rms": 0.16155554713844067, + "teaching_rms": 0.002638740849668689, + "traffic_rms": 0.16153365975217523 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.812056353855951e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6105316625435605e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014472707629248193, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.37117434242248537 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.100070834159851, + "mean_feedback_forward_cosine": 0.7844552457332611, + "mean_feedback_forward_relative_error": 0.6391675168817693, + "min_feedback_forward_cosine": 0.4194483458995819 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026553206617605882, + "instruction_rms": 0.0026553206618531036, + "raw_apical_rms": 0.16272969824856956, + "teaching_rms": 0.0026553206617605882, + "traffic_rms": 0.162707692791046 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8468625739842537e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.604893077380114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014989256939306582, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.36576996119499205 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0814108848571777, + "mean_feedback_forward_cosine": 0.7970635105263103, + "mean_feedback_forward_relative_error": 0.6191378029910001, + "min_feedback_forward_cosine": 0.4391012191772461 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002600648842565709, + "instruction_rms": 0.002600648842670833, + "raw_apical_rms": 0.16438005314662213, + "teaching_rms": 0.002600648842565709, + "traffic_rms": 0.16435911168515158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.520546248912069e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.594728109151353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014997366792962244, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3507835594367981 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.062459111213684, + "mean_feedback_forward_cosine": 0.8090392540801655, + "mean_feedback_forward_relative_error": 0.5996172953735698, + "min_feedback_forward_cosine": 0.4590038061141968 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002588869621380008, + "instruction_rms": 0.0025888696212781736, + "raw_apical_rms": 0.16556999798776575, + "teaching_rms": 0.002588869621380008, + "traffic_rms": 0.16554943689273396 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4964957983920613e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5922249343910863e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015333313823188229, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3443980696773529 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0438786745071411, + "mean_feedback_forward_cosine": 0.8205098162997853, + "mean_feedback_forward_relative_error": 0.5803266414187171, + "min_feedback_forward_cosine": 0.47844213247299194 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026061415885313827, + "instruction_rms": 0.002606141587827657, + "raw_apical_rms": 0.1663770539222545, + "teaching_rms": 0.0026061415885313827, + "traffic_rms": 0.16635627296729813 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 8.985276878092918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585463528392939e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015132403563762123, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.32876510339736936 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0252468585968018, + "mean_feedback_forward_cosine": 0.8310683857310902, + "mean_feedback_forward_relative_error": 0.5619848256761377, + "min_feedback_forward_cosine": 0.4973334074020386 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002594661694146906, + "instruction_rms": 0.00259466169391739, + "raw_apical_rms": 0.16763099888632327, + "teaching_rms": 0.002594661694146906, + "traffic_rms": 0.16761052494835652 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.3831893852511712e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5820250913130137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015345625675080016, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3172031030368805 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0044704675674438, + "mean_feedback_forward_cosine": 0.8417373342947526, + "mean_feedback_forward_relative_error": 0.5431437801231037, + "min_feedback_forward_cosine": 0.518530547618866 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026471006429492024, + "instruction_rms": 0.0026471006437781453, + "raw_apical_rms": 0.1692845885272242, + "teaching_rms": 0.0026471006429492024, + "traffic_rms": 0.16926348651806883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 9.705438230867003e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5844411451785783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015242858073868814, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31501501554489136 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9849327802658081, + "mean_feedback_forward_cosine": 0.8514075127514926, + "mean_feedback_forward_relative_error": 0.5254123882813887, + "min_feedback_forward_cosine": 0.5373629331588745 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002626913143034699, + "instruction_rms": 0.002626913142189967, + "raw_apical_rms": 0.17012702910887942, + "teaching_rms": 0.002626913143034699, + "traffic_rms": 0.1701063649853961 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 1.0083804369287463e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5913865128588114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015556009236562272, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.30516346851348874 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9628252387046814, + "mean_feedback_forward_cosine": 0.8606625676155091, + "mean_feedback_forward_relative_error": 0.5079831020398573, + "min_feedback_forward_cosine": 0.5576643943786621 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002620524644615033, + "instruction_rms": 0.0026205246460973587, + "raw_apical_rms": 0.17090448783003842, + "teaching_rms": 0.002620524644615033, + "traffic_rms": 0.17088399544645952 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 1.243406586581841e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5782523046894798e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01488105975955009, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29595538734912874 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9428328275680542, + "mean_feedback_forward_cosine": 0.869263971935619, + "mean_feedback_forward_relative_error": 0.4912926562807777, + "min_feedback_forward_cosine": 0.5763156414031982 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025896140552726997, + "instruction_rms": 0.0025896140559091853, + "raw_apical_rms": 0.17205904245988815, + "teaching_rms": 0.0025896140552726997, + "traffic_rms": 0.1720391072836168 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 6.793288065409797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.590124014954709e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014743848880003723, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.28293708559036257 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9215667843818665, + "mean_feedback_forward_cosine": 0.8775421424345536, + "mean_feedback_forward_relative_error": 0.4747281459244815, + "min_feedback_forward_cosine": 0.5951628684997559 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026134032244007448, + "instruction_rms": 0.0026134032236832254, + "raw_apical_rms": 0.1729953795666652, + "teaching_rms": 0.0026134032244007448, + "traffic_rms": 0.17297519210044943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 8.425057984595696e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5734427730825037e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014970520019791366, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.28208935883522035 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8995399475097656, + "mean_feedback_forward_cosine": 0.8854209769855846, + "mean_feedback_forward_relative_error": 0.45858497362245215, + "min_feedback_forward_cosine": 0.6142142415046692 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025909411529305613, + "instruction_rms": 0.0025909411522636377, + "raw_apical_rms": 0.1737467853710368, + "teaching_rms": 0.0025909411529305613, + "traffic_rms": 0.1737270865150834 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.248495635763902e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.576603320839748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014703149804685865, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2739486166000366 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8778071403503418, + "mean_feedback_forward_cosine": 0.8927729021419178, + "mean_feedback_forward_relative_error": 0.443053355271166, + "min_feedback_forward_cosine": 0.6325311660766602 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025672040524188115, + "instruction_rms": 0.002567204052023463, + "raw_apical_rms": 0.17445495344267126, + "teaching_rms": 0.0025672040524188115, + "traffic_rms": 0.17443562442690488 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.4424638550281088e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5811056151736797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014509353491457619, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26884742764472963 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8552789092063904, + "mean_feedback_forward_cosine": 0.9000179973515597, + "mean_feedback_forward_relative_error": 0.4273221724412658, + "min_feedback_forward_cosine": 0.6514319777488708 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002622781527678394, + "instruction_rms": 0.002622781526873354, + "raw_apical_rms": 0.17542996504543096, + "teaching_rms": 0.002622781527678394, + "traffic_rms": 0.17540995777953275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.460579426293407e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5748145586684298e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014805080570001624, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.262814901137352 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8326566815376282, + "mean_feedback_forward_cosine": 0.9066807302561674, + "mean_feedback_forward_relative_error": 0.41234598823569035, + "min_feedback_forward_cosine": 0.6695079207420349 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025832089140066026, + "instruction_rms": 0.002583208913886462, + "raw_apical_rms": 0.17591821193988708, + "teaching_rms": 0.0025832089140066026, + "traffic_rms": 0.17589880692646895 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 9.167460035541808e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.578851214323037e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014703941689974425, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.26087324299812314 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8089280724525452, + "mean_feedback_forward_cosine": 0.9130848971280184, + "mean_feedback_forward_relative_error": 0.3974617175080559, + "min_feedback_forward_cosine": 0.6878708600997925 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026129884058324806, + "instruction_rms": 0.0026129884042660587, + "raw_apical_rms": 0.17695388151463512, + "teaching_rms": 0.0026129884058324806, + "traffic_rms": 0.17693412296572925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 9.196924821708308e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5690377181671175e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01536738468614489, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.25386013025283816 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7852631211280823, + "mean_feedback_forward_cosine": 0.9190836722200567, + "mean_feedback_forward_relative_error": 0.38311606699770145, + "min_feedback_forward_cosine": 0.7059113383293152 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025927921948294685, + "instruction_rms": 0.0025927921956414397, + "raw_apical_rms": 0.17769609319234955, + "teaching_rms": 0.0025927921948294685, + "traffic_rms": 0.17767667295761647 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 1.0377942771810922e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5748011782173054e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015187333786801177, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25115535389900207 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7639781832695007, + "mean_feedback_forward_cosine": 0.9244459022175182, + "mean_feedback_forward_relative_error": 0.369787706841122, + "min_feedback_forward_cosine": 0.7214045524597168 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025554021834269624, + "instruction_rms": 0.0025554021830296917, + "raw_apical_rms": 0.1783578825622431, + "teaching_rms": 0.0025554021834269624, + "traffic_rms": 0.17833911107017594 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.1113689879493904e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5654474901864392e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015180681447650834, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.24250418165206908 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7412458062171936, + "mean_feedback_forward_cosine": 0.9297022982077165, + "mean_feedback_forward_relative_error": 0.3563449056311087, + "min_feedback_forward_cosine": 0.7376618981361389 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025981334355545567, + "instruction_rms": 0.002598133435323027, + "raw_apical_rms": 0.17861867997734054, + "teaching_rms": 0.0025981334355545567, + "traffic_rms": 0.17859931400400342 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.3657998465532728e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.578146292293606e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01492229117337729, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.2394808725976944 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7192479968070984, + "mean_feedback_forward_cosine": 0.934744492444125, + "mean_feedback_forward_relative_error": 0.34301622699607504, + "min_feedback_forward_cosine": 0.7525535821914673 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002606824459723985, + "instruction_rms": 0.002606824460028297, + "raw_apical_rms": 0.17967835746277985, + "teaching_rms": 0.002606824459723985, + "traffic_rms": 0.1796589641850412 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.2859233322615182e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.571678717373791e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014919413944707131, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2386718812561035 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6980090141296387, + "mean_feedback_forward_cosine": 0.9392488132823598, + "mean_feedback_forward_relative_error": 0.330627683753317, + "min_feedback_forward_cosine": 0.7669057250022888 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026090958875849384, + "instruction_rms": 0.0026090958886405106, + "raw_apical_rms": 0.18103805839634887, + "teaching_rms": 0.0026090958875849384, + "traffic_rms": 0.18101873344022157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 8.5686941986296e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.564222478124667e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015138517281139213, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23023530738830567 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6771202683448792, + "mean_feedback_forward_cosine": 0.9435265985402194, + "mean_feedback_forward_relative_error": 0.31856826476075434, + "min_feedback_forward_cosine": 0.7804913520812988 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025699343460919167, + "instruction_rms": 0.002569934346407039, + "raw_apical_rms": 0.18186836538008372, + "teaching_rms": 0.0025699343460919167, + "traffic_rms": 0.18184970888757424 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032893240451813, + "maximum_absolute_post_projection_soma_slope": 6.367667282347611e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5638642806633304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015165605709411209, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22689663739204408 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6567354798316956, + "mean_feedback_forward_cosine": 0.9474854642694647, + "mean_feedback_forward_relative_error": 0.306992983547124, + "min_feedback_forward_cosine": 0.793333113193512 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025823542212689947, + "instruction_rms": 0.0025823542215574983, + "raw_apical_rms": 0.18210764089370307, + "teaching_rms": 0.0025823542212689947, + "traffic_rms": 0.18208877856758035 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 9.688846347444269e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5568095251289484e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015120585329014342, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22847433372497558 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6365740299224854, + "mean_feedback_forward_cosine": 0.9512472867965698, + "mean_feedback_forward_relative_error": 0.2955930929292332, + "min_feedback_forward_cosine": 0.8056755661964417 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002588035387629837, + "instruction_rms": 0.002588035387692982, + "raw_apical_rms": 0.18307689246159592, + "teaching_rms": 0.002588035387629837, + "traffic_rms": 0.18305820097162234 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.10480343096242e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5574105573045435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01452721422828918, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22121686527252196 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6166413426399231, + "mean_feedback_forward_cosine": 0.9546859914606268, + "mean_feedback_forward_relative_error": 0.2847704238512299, + "min_feedback_forward_cosine": 0.81741863489151 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002548038506011622, + "instruction_rms": 0.002548038506014367, + "raw_apical_rms": 0.18450616634386022, + "teaching_rms": 0.002548038506011622, + "traffic_rms": 0.18448803823301446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.5567027844554104e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5537693321807416e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014057093535507095, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.21667400578022003 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5959099531173706, + "mean_feedback_forward_cosine": 0.9579202110117132, + "mean_feedback_forward_relative_error": 0.2743028810078448, + "min_feedback_forward_cosine": 0.8293538689613342 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002541597892012647, + "instruction_rms": 0.0025415978920622375, + "raw_apical_rms": 0.18508047805223146, + "teaching_rms": 0.002541597892012647, + "traffic_rms": 0.18506255146405443 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6250916701210372e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.546305044600369e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014706828258071025, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.21273947812080382 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5753595232963562, + "mean_feedback_forward_cosine": 0.9609669977968389, + "mean_feedback_forward_relative_error": 0.26406141113151205, + "min_feedback_forward_cosine": 0.8407213687896729 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025764539219527262, + "instruction_rms": 0.0025764539218346297, + "raw_apical_rms": 0.1857082439262729, + "teaching_rms": 0.0025764539219527262, + "traffic_rms": 0.18568991083097625 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5844125300645828e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5496568644682882e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014986624402248816, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21046549332141876 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5565568208694458, + "mean_feedback_forward_cosine": 0.9637275316498496, + "mean_feedback_forward_relative_error": 0.2544814726168459, + "min_feedback_forward_cosine": 0.85085129737854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025214766910815834, + "instruction_rms": 0.0025214766904931725, + "raw_apical_rms": 0.18675312203675049, + "teaching_rms": 0.0025214766910815834, + "traffic_rms": 0.1867356387196873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 9.456736904667196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5404349064264783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015439841192754251, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.20768615818977357 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5371419787406921, + "mean_feedback_forward_cosine": 0.9663732517849315, + "mean_feedback_forward_relative_error": 0.24491879357533022, + "min_feedback_forward_cosine": 0.8609918355941772 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025628483246883257, + "instruction_rms": 0.002562848325888633, + "raw_apical_rms": 0.18694774064645125, + "teaching_rms": 0.0025628483246883257, + "traffic_rms": 0.18692968404872878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.4701700479236024e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5471555378220515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014680919902981563, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20576488817453384 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5187984108924866, + "mean_feedback_forward_cosine": 0.9688193537972191, + "mean_feedback_forward_relative_error": 0.23577449057589878, + "min_feedback_forward_cosine": 0.8702529668807983 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002528627383329315, + "instruction_rms": 0.002528627383619348, + "raw_apical_rms": 0.1870826849451422, + "teaching_rms": 0.002528627383329315, + "traffic_rms": 0.18706505180466743 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5776669215483707e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5434519736914357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014529707264139251, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.20054735531806947 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.49979838728904724, + "mean_feedback_forward_cosine": 0.9711570165374063, + "mean_feedback_forward_relative_error": 0.22674356604164297, + "min_feedback_forward_cosine": 0.8794181942939758 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025601900207233353, + "instruction_rms": 0.0025601900209007433, + "raw_apical_rms": 0.18812676793884092, + "teaching_rms": 0.0025601900207233353, + "traffic_rms": 0.18810884284488646 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.9960663166784798e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5467231772023124e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015159105548384762, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.19832884444236756 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48140284419059753, + "mean_feedback_forward_cosine": 0.9733218388123945, + "mean_feedback_forward_relative_error": 0.2180432064966722, + "min_feedback_forward_cosine": 0.8880386352539062 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002589216664413249, + "instruction_rms": 0.0025892166640966055, + "raw_apical_rms": 0.18868869347341313, + "teaching_rms": 0.002589216664413249, + "traffic_rms": 0.1886704403324625 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4819390514730912e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5428403750472704e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015325440061047533, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.2035802125263214 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46508845686912537, + "mean_feedback_forward_cosine": 0.9752543893727389, + "mean_feedback_forward_relative_error": 0.20989659889177842, + "min_feedback_forward_cosine": 0.8954968452453613 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002538441247691544, + "instruction_rms": 0.0025384412475367916, + "raw_apical_rms": 0.18961838824125493, + "teaching_rms": 0.002538441247691544, + "traffic_rms": 0.18960091582516456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.5328072322517983e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54800218779403e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015118734844508838, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.19359651690483093 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4472070634365082, + "mean_feedback_forward_cosine": 0.9771620598706332, + "mean_feedback_forward_relative_error": 0.20167528025128625, + "min_feedback_forward_cosine": 0.9032673239707947 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025743985226011267, + "instruction_rms": 0.0025743985225824186, + "raw_apical_rms": 0.1900623174943143, + "teaching_rms": 0.0025743985226011267, + "traffic_rms": 0.19004435498777772 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2530840365343465e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547547823033396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015310326037885793, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.1995827917098999 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43161067366600037, + "mean_feedback_forward_cosine": 0.978813886642456, + "mean_feedback_forward_relative_error": 0.19421078244393522, + "min_feedback_forward_cosine": 0.9098244905471802 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00252313460787923, + "instruction_rms": 0.0025231346083171816, + "raw_apical_rms": 0.19079279275648747, + "teaching_rms": 0.00252313460787923, + "traffic_rms": 0.19077561797566422 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 9.481016860490854e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5423461180889964e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01595027196058705, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.1954041485786438 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41619226336479187, + "mean_feedback_forward_cosine": 0.9803691387176514, + "mean_feedback_forward_relative_error": 0.18693822513927114, + "min_feedback_forward_cosine": 0.9160608053207397 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002488668378534587, + "instruction_rms": 0.002488668378894413, + "raw_apical_rms": 0.19168626762021632, + "teaching_rms": 0.002488668378534587, + "traffic_rms": 0.1916696529593521 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.0725605648076453e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54772740177661e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015438449396487372, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.1849528354358673 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4006912112236023, + "mean_feedback_forward_cosine": 0.9818727601658214, + "mean_feedback_forward_relative_error": 0.17961767999963327, + "min_feedback_forward_cosine": 0.9220747351646423 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002572890593158409, + "instruction_rms": 0.002572890593217565, + "raw_apical_rms": 0.1923844357598027, + "teaching_rms": 0.002572890593158409, + "traffic_rms": 0.19236670769340591 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2435226892648643e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5401233610444186e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015592495159290521, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.18668059752464294 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3856959640979767, + "mean_feedback_forward_cosine": 0.9832323757084933, + "mean_feedback_forward_relative_error": 0.17273433452302758, + "min_feedback_forward_cosine": 0.927733838558197 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002567611580168549, + "instruction_rms": 0.002567611579041464, + "raw_apical_rms": 0.19356638553968591, + "teaching_rms": 0.002567611580168549, + "traffic_rms": 0.19354880485507386 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.554649458057611e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5457124075594227e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015017205940649928, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.18320563239097595 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.37126052379608154, + "mean_feedback_forward_cosine": 0.9845116658644243, + "mean_feedback_forward_relative_error": 0.16601317910985514, + "min_feedback_forward_cosine": 0.9329911470413208 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025606266291857507, + "instruction_rms": 0.0025606266290862617, + "raw_apical_rms": 0.1940135362142242, + "teaching_rms": 0.0025606266291857507, + "traffic_rms": 0.1939960913559153 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4965364414365467e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544349763126263e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014839039049141694, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1894313395023346 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3578231632709503, + "mean_feedback_forward_cosine": 0.9856740127910267, + "mean_feedback_forward_relative_error": 0.15965098616751758, + "min_feedback_forward_cosine": 0.9376983046531677 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025416139504683645, + "instruction_rms": 0.0025416139500036537, + "raw_apical_rms": 0.1942435985683798, + "teaching_rms": 0.0025416139504683645, + "traffic_rms": 0.19422640580789513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.0029463481942003e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5394075189918815e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014559005914787789, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.1801194595718384 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3451382517814636, + "mean_feedback_forward_cosine": 0.9867381442676891, + "mean_feedback_forward_relative_error": 0.153586978871714, + "min_feedback_forward_cosine": 0.9419865608215332 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025351986737312242, + "instruction_rms": 0.00253519867385464, + "raw_apical_rms": 0.19547401330571518, + "teaching_rms": 0.0025351986737312242, + "traffic_rms": 0.19545702993269928 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.323587213164501e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5415012422517068e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014596486846079167, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.17846159574508666 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3316519856452942, + "mean_feedback_forward_cosine": 0.987756482037631, + "mean_feedback_forward_relative_error": 0.14757510830055584, + "min_feedback_forward_cosine": 0.9463814496994019 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025497222081453568, + "instruction_rms": 0.002549722208627574, + "raw_apical_rms": 0.19561517774541148, + "teaching_rms": 0.0025497222081453568, + "traffic_rms": 0.19559799253814317 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.1047026049482156e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5434148451441177e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0147411236059591, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1797970705318451 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31864625215530396, + "mean_feedback_forward_cosine": 0.9887095581401478, + "mean_feedback_forward_relative_error": 0.14172076427123764, + "min_feedback_forward_cosine": 0.9504607915878296 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026138272934654763, + "instruction_rms": 0.002613827293341555, + "raw_apical_rms": 0.19655806047738084, + "teaching_rms": 0.0026138272934654763, + "traffic_rms": 0.1965401228720232 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.2383709702135093e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5377333435046974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015352866685329985, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18204149487018584 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3065074682235718, + "mean_feedback_forward_cosine": 0.9895738374103199, + "mean_feedback_forward_relative_error": 0.13619210279800675, + "min_feedback_forward_cosine": 0.9541308879852295 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002572100262802312, + "instruction_rms": 0.002572100262922261, + "raw_apical_rms": 0.19732952215422966, + "teaching_rms": 0.002572100262802312, + "traffic_rms": 0.19731216766067375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.4507084244996804e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5353603041954792e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015339630945449715, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.17835984148979186 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2944309413433075, + "mean_feedback_forward_cosine": 0.990371642329476, + "mean_feedback_forward_relative_error": 0.13089491772380743, + "min_feedback_forward_cosine": 0.9576348662376404 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002554533338884294, + "instruction_rms": 0.0025545333389234695, + "raw_apical_rms": 0.1973558530497234, + "teaching_rms": 0.002554533338884294, + "traffic_rms": 0.19733874936590917 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.1720576509333114e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5483868134959536e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015569585591090854, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.17530373861312867 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2834106981754303, + "mean_feedback_forward_cosine": 0.9911048206416043, + "mean_feedback_forward_relative_error": 0.1258192678066817, + "min_feedback_forward_cosine": 0.9607114791870117 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025746487242735558, + "instruction_rms": 0.002574648724071054, + "raw_apical_rms": 0.19807789914100749, + "teaching_rms": 0.0025746487242735558, + "traffic_rms": 0.1980605922225458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.499005435334766e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5370463661084006e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015097284354703715, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17049965922355653 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27217045426368713, + "mean_feedback_forward_cosine": 0.9918010473251343, + "mean_feedback_forward_relative_error": 0.12083019329742951, + "min_feedback_forward_cosine": 0.9637404680252075 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025844669424087796, + "instruction_rms": 0.0025844669415708683, + "raw_apical_rms": 0.1991234781767051, + "teaching_rms": 0.0025844669424087796, + "traffic_rms": 0.1991061838658178 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3173192314752669e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54070652371076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01531612643605192, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.17079505893707275 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.261589378118515, + "mean_feedback_forward_cosine": 0.9924263076348738, + "mean_feedback_forward_relative_error": 0.11615983298556372, + "min_feedback_forward_cosine": 0.9664864540100098 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002553857736284652, + "instruction_rms": 0.0025538577366070255, + "raw_apical_rms": 0.19921165696005638, + "teaching_rms": 0.002553857736284652, + "traffic_rms": 0.19919467874578778 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.379805212309293e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5470375282357076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015038088872024243, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.169794254655838 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2512308359146118, + "mean_feedback_forward_cosine": 0.9929907191883434, + "mean_feedback_forward_relative_error": 0.11179501590403644, + "min_feedback_forward_cosine": 0.9690577387809753 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025338126234168376, + "instruction_rms": 0.0025338126243056903, + "raw_apical_rms": 0.1997061723815295, + "teaching_rms": 0.0025338126234168376, + "traffic_rms": 0.19968952336306153 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.1868691274230514e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5399078728928208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015532350836118595, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.16322729253768922 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24170196056365967, + "mean_feedback_forward_cosine": 0.9935119076208635, + "mean_feedback_forward_relative_error": 0.10757296346127987, + "min_feedback_forward_cosine": 0.9713436365127563 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002497489869652494, + "instruction_rms": 0.0024974898699108316, + "raw_apical_rms": 0.20118509556302, + "teaching_rms": 0.002497489869652494, + "traffic_rms": 0.20116906579103574 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4423747529690445e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5451224458784764e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015600147415412682, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.1631049171566963 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2328723967075348, + "mean_feedback_forward_cosine": 0.9940059466795488, + "mean_feedback_forward_relative_error": 0.10341610349714755, + "min_feedback_forward_cosine": 0.9733949303627014 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025335086264244404, + "instruction_rms": 0.002533508626302085, + "raw_apical_rms": 0.20116239337587105, + "teaching_rms": 0.0025335086264244404, + "traffic_rms": 0.20114576396101447 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5484329196624458e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5430856812356932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015207227133053454, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.16361883125543594 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22352224588394165, + "mean_feedback_forward_cosine": 0.994468979402022, + "mean_feedback_forward_relative_error": 0.0993709310889244, + "min_feedback_forward_cosine": 0.975462019443512 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025574356725624058, + "instruction_rms": 0.0025574356726721127, + "raw_apical_rms": 0.20206905183455764, + "teaching_rms": 0.0025574356725624058, + "traffic_rms": 0.2020522444855574 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3651930430569337e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5539378051753855e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015396487332990485, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16296307560920714 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21477417647838593, + "mean_feedback_forward_cosine": 0.994889783859253, + "mean_feedback_forward_relative_error": 0.09553489739244635, + "min_feedback_forward_cosine": 0.9773380160331726 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002531463878518537, + "instruction_rms": 0.0025314638781263406, + "raw_apical_rms": 0.20306262273214315, + "teaching_rms": 0.002531463878518537, + "traffic_rms": 0.20304627503901643 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.171533341448594e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5490857810413483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015934908475166366, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.16344164234161376 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2060430943965912, + "mean_feedback_forward_cosine": 0.9952837727286599, + "mean_feedback_forward_relative_error": 0.09180980121547526, + "min_feedback_forward_cosine": 0.9791228771209717 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025606623697269847, + "instruction_rms": 0.0025606623702008505, + "raw_apical_rms": 0.20308522750543848, + "teaching_rms": 0.0025606623697269847, + "traffic_rms": 0.20306853010358134 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.7227115733930987e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5497857664108928e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016396496003387637, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16472641209602357 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19783030450344086, + "mean_feedback_forward_cosine": 0.9956494613127275, + "mean_feedback_forward_relative_error": 0.08819995827295564, + "min_feedback_forward_cosine": 0.9807487726211548 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002584683773817322, + "instruction_rms": 0.0025846837733310883, + "raw_apical_rms": 0.20395892595937332, + "teaching_rms": 0.002584683773817322, + "traffic_rms": 0.20394196469665066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2110133695841796e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5484886437441268e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015480462488414592, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16252612312793732 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1904464066028595, + "mean_feedback_forward_cosine": 0.9959717674688859, + "mean_feedback_forward_relative_error": 0.0848703970624642, + "min_feedback_forward_cosine": 0.982141375541687 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025193337403480722, + "instruction_rms": 0.0025193337409264204, + "raw_apical_rms": 0.20455933355472664, + "teaching_rms": 0.0025193337403480722, + "traffic_rms": 0.2045432527317793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6840667171891255e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5461793983975394e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01568088890836231, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.15821564504146576 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18275530636310577, + "mean_feedback_forward_cosine": 0.9962850874120539, + "mean_feedback_forward_relative_error": 0.081526034664024, + "min_feedback_forward_cosine": 0.9835509657859802 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002554171861005083, + "instruction_rms": 0.0025541718610707356, + "raw_apical_rms": 0.20566982598017788, + "teaching_rms": 0.002554171861005083, + "traffic_rms": 0.20565335125330692 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3324543601811456e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5466280423285437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01644930075499787, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.16145968150138856 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17528261244297028, + "mean_feedback_forward_cosine": 0.9965692541816018, + "mean_feedback_forward_relative_error": 0.07836330258710818, + "min_feedback_forward_cosine": 0.9848649501800537 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002520254653503778, + "instruction_rms": 0.002520254653841636, + "raw_apical_rms": 0.2061784808327215, + "teaching_rms": 0.002520254653503778, + "traffic_rms": 0.20616245740672506 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.8999432427335705e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.55117969229901e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016021993095052745, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.15601355092287064 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16802464425563812, + "mean_feedback_forward_cosine": 0.9968333005905151, + "mean_feedback_forward_relative_error": 0.07532467713410204, + "min_feedback_forward_cosine": 0.986082136631012 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002521226693617987, + "instruction_rms": 0.0025212266935479488, + "raw_apical_rms": 0.20623155824144473, + "teaching_rms": 0.002521226693617987, + "traffic_rms": 0.20621555485104795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.6114196910166356e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5508889667814607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016006569143387606, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15837253064632414 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16154427826404572, + "mean_feedback_forward_cosine": 0.9970792445269498, + "mean_feedback_forward_relative_error": 0.07235202755440365, + "min_feedback_forward_cosine": 0.9871301054954529 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025787924648255998, + "instruction_rms": 0.0025787924656251056, + "raw_apical_rms": 0.20641577146954043, + "teaching_rms": 0.0025787924648255998, + "traffic_rms": 0.20639903956726888 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.1262209653750688e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5456007033959386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015705127146941278, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15660233090877532 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15518537163734436, + "mean_feedback_forward_cosine": 0.997306210344488, + "mean_feedback_forward_relative_error": 0.06950314955955202, + "min_feedback_forward_cosine": 0.9881171584129333 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00256035335132143, + "instruction_rms": 0.00256035335126586, + "raw_apical_rms": 0.20657903224248886, + "teaching_rms": 0.00256035335132143, + "traffic_rms": 0.20656257388476035 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8357698650106613e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5457154641675084e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015514368554078903, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15804204572200775 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14864493906497955, + "mean_feedback_forward_cosine": 0.9975167935544794, + "mean_feedback_forward_relative_error": 0.06676234738393264, + "min_feedback_forward_cosine": 0.9890908002853394 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002574814127116803, + "instruction_rms": 0.002574814126784988, + "raw_apical_rms": 0.20729438777520215, + "teaching_rms": 0.002574814127116803, + "traffic_rms": 0.20727776252416127 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2181038755443296e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5377016392669557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015869710780527502, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15755722439289094 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14300605654716492, + "mean_feedback_forward_cosine": 0.9977021759206598, + "mean_feedback_forward_relative_error": 0.06423464827239514, + "min_feedback_forward_cosine": 0.9898954033851624 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025335077956092425, + "instruction_rms": 0.002533507795517158, + "raw_apical_rms": 0.20732922200601905, + "teaching_rms": 0.0025335077956092425, + "traffic_rms": 0.20731317192770435 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7502101457012031e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5392315794784796e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015584259231916607, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1498199157810211 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13754555583000183, + "mean_feedback_forward_cosine": 0.99787705486471, + "mean_feedback_forward_relative_error": 0.06175127991221168, + "min_feedback_forward_cosine": 0.990649938583374 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002551391654005273, + "instruction_rms": 0.0025513916538489672, + "raw_apical_rms": 0.20810134466862668, + "teaching_rms": 0.002551391654005273, + "traffic_rms": 0.20808503615334314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.568492962178425e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5441517916104492e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015558901363756272, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15399860085010528 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13223497569561005, + "mean_feedback_forward_cosine": 0.9980346202850342, + "mean_feedback_forward_relative_error": 0.05942273439670151, + "min_feedback_forward_cosine": 0.9913565516471863 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002513373221570627, + "instruction_rms": 0.002513373221535654, + "raw_apical_rms": 0.2088181781539231, + "teaching_rms": 0.002513373221570627, + "traffic_rms": 0.20880239633232103 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 2.0101597897337342e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544821265609005e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01576733974504039, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.14976577283859252 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12683920562267303, + "mean_feedback_forward_cosine": 0.9981916449286721, + "mean_feedback_forward_relative_error": 0.05702021513134241, + "min_feedback_forward_cosine": 0.9920450448989868 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026000874852045914, + "instruction_rms": 0.0026000874836015543, + "raw_apical_rms": 0.20902082608472825, + "teaching_rms": 0.0026000874852045914, + "traffic_rms": 0.20900397635114182 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.808545277981466e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5379354044835517e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01563514234409263, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1550193964147568 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12157892435789108, + "mean_feedback_forward_cosine": 0.9983321753415194, + "mean_feedback_forward_relative_error": 0.054779926873743535, + "min_feedback_forward_cosine": 0.9926872253417969 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025827614770866287, + "instruction_rms": 0.0025827614767556816, + "raw_apical_rms": 0.20970305803554404, + "teaching_rms": 0.0025827614770866287, + "traffic_rms": 0.20968656414630513 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 2.2467528992820007e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5434473403663686e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015514940885966843, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15298853549957275 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11665408313274384, + "mean_feedback_forward_cosine": 0.998455781286413, + "mean_feedback_forward_relative_error": 0.05273575789549134, + "min_feedback_forward_cosine": 0.9932650327682495 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002499096932659557, + "instruction_rms": 0.0024990969319226264, + "raw_apical_rms": 0.2106382994830396, + "teaching_rms": 0.002499096932659557, + "traffic_rms": 0.21062290346834134 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4873180020913424e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545161806962649e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015482689101691875, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.14483871377944946 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11181838065385818, + "mean_feedback_forward_cosine": 0.9985774983059276, + "mean_feedback_forward_relative_error": 0.05063163285905665, + "min_feedback_forward_cosine": 0.9938100576400757 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002599968539224205, + "instruction_rms": 0.002599968539744747, + "raw_apical_rms": 0.21118512039618645, + "teaching_rms": 0.002599968539224205, + "traffic_rms": 0.21116853335694002 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 9.493436436969205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5521455423873283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015662692255687543, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.15348670791625976 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1071416437625885, + "mean_feedback_forward_cosine": 0.9986843976107511, + "mean_feedback_forward_relative_error": 0.04870991142974659, + "min_feedback_forward_cosine": 0.9943141341209412 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025186034280714168, + "instruction_rms": 0.0025186034282199932, + "raw_apical_rms": 0.21144440611651413, + "teaching_rms": 0.0025186034280714168, + "traffic_rms": 0.2114288112370692 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5013283416465129e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.548217615560487e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016331211646262694, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14709635178685188 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10299846529960632, + "mean_feedback_forward_cosine": 0.9987838831814853, + "mean_feedback_forward_relative_error": 0.046841320598667316, + "min_feedback_forward_cosine": 0.9947429895401001 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025452271584177556, + "instruction_rms": 0.002545227159053103, + "raw_apical_rms": 0.21216240491789193, + "teaching_rms": 0.0025452271584177556, + "traffic_rms": 0.21214646854293032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2457590514713957e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5481031038492368e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015799631922532087, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.149340011305809 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09870495647192001, + "mean_feedback_forward_cosine": 0.9988782048225403, + "mean_feedback_forward_relative_error": 0.045013465715402906, + "min_feedback_forward_cosine": 0.9951716661453247 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025365940414007896, + "instruction_rms": 0.002536594041109598, + "raw_apical_rms": 0.21228205727383279, + "teaching_rms": 0.0025365940414007896, + "traffic_rms": 0.2122663947019807 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4391649472145218e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5378829029443448e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015714520130114963, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1438264095211029 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0947856456041336, + "mean_feedback_forward_cosine": 0.9989660599014976, + "mean_feedback_forward_relative_error": 0.04322832272472707, + "min_feedback_forward_cosine": 0.995545506477356 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025645730509414903, + "instruction_rms": 0.0025645730511021972, + "raw_apical_rms": 0.21285123196543732, + "teaching_rms": 0.0025645730509414903, + "traffic_rms": 0.21283511828524135 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3644201146689738e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5460083714942875e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0163829156348411, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.14436068420886994 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09082983434200287, + "mean_feedback_forward_cosine": 0.9990450555627997, + "mean_feedback_forward_relative_error": 0.041559927182441406, + "min_feedback_forward_cosine": 0.9959096312522888 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025547090209680047, + "instruction_rms": 0.002554709021331771, + "raw_apical_rms": 0.21337583584845468, + "teaching_rms": 0.0025547090209680047, + "traffic_rms": 0.21335995854587772 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.1935662769246846e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.548143668937137e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015794205791579155, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.1461905869436264 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08700595796108246, + "mean_feedback_forward_cosine": 0.9991198149594394, + "mean_feedback_forward_relative_error": 0.03991693876345049, + "min_feedback_forward_cosine": 0.9962445497512817 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025941327591904634, + "instruction_rms": 0.002594132759417947, + "raw_apical_rms": 0.2138526598530522, + "teaching_rms": 0.0025941327591904634, + "traffic_rms": 0.2138362680871085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.3270846466184594e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.553065769989632e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016122140670168858, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14529388692378997 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08350390195846558, + "mean_feedback_forward_cosine": 0.9991890874775973, + "mean_feedback_forward_relative_error": 0.038322304290803993, + "min_feedback_forward_cosine": 0.9965389966964722 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026024407996732735, + "instruction_rms": 0.002602440800322515, + "raw_apical_rms": 0.21484563487598907, + "teaching_rms": 0.0026024407996732735, + "traffic_rms": 0.2148292440446389 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.515123244644201e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5571430350303438e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01619996466637663, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14788773932933807 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08038336038589478, + "mean_feedback_forward_cosine": 0.999249821359461, + "mean_feedback_forward_relative_error": 0.036867358061400325, + "min_feedback_forward_cosine": 0.9967923760414124 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002551340353409675, + "instruction_rms": 0.002551340353525551, + "raw_apical_rms": 0.21548983513705708, + "teaching_rms": 0.002551340353409675, + "traffic_rms": 0.21547412852202397 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5439940170836053e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5396404312287556e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015404332962239213, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.14296333221435548 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07727920264005661, + "mean_feedback_forward_cosine": 0.9993063298138705, + "mean_feedback_forward_relative_error": 0.03546306028623473, + "min_feedback_forward_cosine": 0.9970346689224243 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025664876842481513, + "instruction_rms": 0.0025664876839330743, + "raw_apical_rms": 0.2162780228746105, + "teaching_rms": 0.0025664876842481513, + "traffic_rms": 0.21626215768836857 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.393754445189188e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543916438311291e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015439747075489115, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.13880622097492218 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07403521984815598, + "mean_feedback_forward_cosine": 0.999360838803378, + "mean_feedback_forward_relative_error": 0.03405545497821136, + "min_feedback_forward_cosine": 0.9972783923149109 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002597284787644532, + "instruction_rms": 0.0025972847876476173, + "raw_apical_rms": 0.21663813092745565, + "teaching_rms": 0.002597284787644532, + "traffic_rms": 0.21662188337039306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.1445144565414012e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5528905908744045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014992794369437144, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14473648000717163 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0709739699959755, + "mean_feedback_forward_cosine": 0.9994099790399725, + "mean_feedback_forward_relative_error": 0.03272983858531171, + "min_feedback_forward_cosine": 0.9974974989891052 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025421352832259675, + "instruction_rms": 0.002542135283591132, + "raw_apical_rms": 0.2174098647825082, + "teaching_rms": 0.0025421352832259675, + "traffic_rms": 0.21739444200614005 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8534331047703745e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.546216377052214e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015235967441663313, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.14428532707214356 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06808094680309296, + "mean_feedback_forward_cosine": 0.9994550477374684, + "mean_feedback_forward_relative_error": 0.03147118657150052, + "min_feedback_forward_cosine": 0.9976972341537476 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025400407868642443, + "instruction_rms": 0.002540040786614386, + "raw_apical_rms": 0.21805662018861477, + "teaching_rms": 0.0025400407868642443, + "traffic_rms": 0.21804120988649228 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7127099738445395e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545722142855208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01539082399721877, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.13967401676654817 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06526024639606476, + "mean_feedback_forward_cosine": 0.9994984117421237, + "mean_feedback_forward_relative_error": 0.030203275331719355, + "min_feedback_forward_cosine": 0.9978837966918945 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025995782730815513, + "instruction_rms": 0.0025995782737711854, + "raw_apical_rms": 0.21820465185257623, + "teaching_rms": 0.0025995782730815513, + "traffic_rms": 0.2181885694437238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6579251393977756e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5446910266130172e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015063445742017702, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.14460846729755403 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06523686647415161, + "mean_feedback_forward_cosine": 0.9994987585327842, + "mean_feedback_forward_relative_error": 0.030192138440907002, + "min_feedback_forward_cosine": 0.9978854060173035 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001801360489677352, + "instruction_rms": 0.0018013604893849962, + "raw_apical_rms": 0.2185080309605919, + "teaching_rms": 0.001801360489677352, + "traffic_rms": 0.21850009727702607 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.3405420418166614e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53581055286684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014309447461897377, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.0765202654504776 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06522545218467712, + "mean_feedback_forward_cosine": 0.9994989254257896, + "mean_feedback_forward_relative_error": 0.030186215813525697, + "min_feedback_forward_cosine": 0.9978863000869751 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015749514162713442, + "instruction_rms": 0.0015749514160202265, + "raw_apical_rms": 0.21861533631350644, + "teaching_rms": 0.0015749514162713442, + "traffic_rms": 0.21860911866633434 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.577511881123428e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5383270491589017e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014361145953508224, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.051795715140104295 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06521355360746384, + "mean_feedback_forward_cosine": 0.9994991009885614, + "mean_feedback_forward_relative_error": 0.03018070703203028, + "min_feedback_forward_cosine": 0.9978870749473572 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015160467936252506, + "instruction_rms": 0.0015160467933003187, + "raw_apical_rms": 0.21869912036959088, + "teaching_rms": 0.0015160467936252506, + "traffic_rms": 0.21869328610706265 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4126845826467616e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545190858461231e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01441803012769257, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04327250795483589 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0652022436261177, + "mean_feedback_forward_cosine": 0.9994992635466836, + "mean_feedback_forward_relative_error": 0.030175569695843892, + "min_feedback_forward_cosine": 0.9978879690170288 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014497995474591853, + "instruction_rms": 0.0014497995473448755, + "raw_apical_rms": 0.21860605203444777, + "teaching_rms": 0.0014497995474591853, + "traffic_rms": 0.2186006189016385 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.873474815283771e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539847568561718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014412972238429137, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.03643689537167549 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06519048660993576, + "mean_feedback_forward_cosine": 0.9994993881745772, + "mean_feedback_forward_relative_error": 0.030170499254018067, + "min_feedback_forward_cosine": 0.9978888034820557 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014167587743835535, + "instruction_rms": 0.001416758773754449, + "raw_apical_rms": 0.21848080917101984, + "teaching_rms": 0.0014167587743835535, + "traffic_rms": 0.21847551272459179 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.37927244736602e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541668096270929e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014502924313056451, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.0330510966026783 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06517927348613739, + "mean_feedback_forward_cosine": 0.9994995388117703, + "mean_feedback_forward_relative_error": 0.030165598837828093, + "min_feedback_forward_cosine": 0.9978895783424377 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001384434579091538, + "instruction_rms": 0.001384434579848739, + "raw_apical_rms": 0.2184254630030536, + "teaching_rms": 0.001384434579091538, + "traffic_rms": 0.21842034477623426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.254165056252532e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5409754979123856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014519246930139128, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.03017406428039074 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06516718864440918, + "mean_feedback_forward_cosine": 0.9994997067884965, + "mean_feedback_forward_relative_error": 0.030160697778178886, + "min_feedback_forward_cosine": 0.9978903532028198 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00133916500905759, + "instruction_rms": 0.0013391650088899307, + "raw_apical_rms": 0.2184108708424228, + "teaching_rms": 0.00133916500905759, + "traffic_rms": 0.2184059005765661 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.7838961241523066e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5458092141008194e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01443608572717173, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.026664466739594935 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06515530496835709, + "mean_feedback_forward_cosine": 0.999499837918715, + "mean_feedback_forward_relative_error": 0.030155824979936536, + "min_feedback_forward_cosine": 0.9978913068771362 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001345419659562782, + "instruction_rms": 0.0013454196590574792, + "raw_apical_rms": 0.2184233548343658, + "teaching_rms": 0.001345419659562782, + "traffic_rms": 0.2184183853964157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3273127308366384e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5424924042122578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014430909533643113, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.024662187173962593 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0651431530714035, + "mean_feedback_forward_cosine": 0.9994999939745123, + "mean_feedback_forward_relative_error": 0.03015116409144618, + "min_feedback_forward_cosine": 0.9978920221328735 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013213742219267057, + "instruction_rms": 0.0013213742213961832, + "raw_apical_rms": 0.21841135438031303, + "teaching_rms": 0.0013213742219267057, + "traffic_rms": 0.21840651912473863 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.5876952375037945e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5393250310448105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014451256823408426, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.023348271403610708 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06513062119483948, + "mean_feedback_forward_cosine": 0.9995001391931013, + "mean_feedback_forward_relative_error": 0.030146537501026284, + "min_feedback_forward_cosine": 0.9978928565979004 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012768604152037176, + "instruction_rms": 0.0012768604152381257, + "raw_apical_rms": 0.21837043748769602, + "teaching_rms": 0.0012768604152037176, + "traffic_rms": 0.2183658239334037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.6456527873742743e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5430403397128237e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014583312963481955, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.021489043878912924 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06511913985013962, + "mean_feedback_forward_cosine": 0.9995002768256448, + "mean_feedback_forward_relative_error": 0.03014194964515892, + "min_feedback_forward_cosine": 0.9978936314582825 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012421149544280395, + "instruction_rms": 0.001242114954396297, + "raw_apical_rms": 0.21827582929154282, + "teaching_rms": 0.0012421149544280395, + "traffic_rms": 0.21827130566555591 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3725160386002244e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5427656033103478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014645267959485665, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.01970999290138483 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0651070699095726, + "mean_feedback_forward_cosine": 0.9995004285465587, + "mean_feedback_forward_relative_error": 0.030137285446240142, + "min_feedback_forward_cosine": 0.9978944659233093 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012560909557728197, + "instruction_rms": 0.0012560909560393393, + "raw_apical_rms": 0.21822951174402053, + "teaching_rms": 0.0012560909557728197, + "traffic_rms": 0.21822500228803843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.9287223551600619e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5436386902386197e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014752025505846596, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.018547217898368835 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06509517878293991, + "mean_feedback_forward_cosine": 0.9995005802674727, + "mean_feedback_forward_relative_error": 0.030132920201867818, + "min_feedback_forward_cosine": 0.9978952407836914 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011600505141997962, + "instruction_rms": 0.001160050514225723, + "raw_apical_rms": 0.21827510921193974, + "teaching_rms": 0.0011600505141997962, + "traffic_rms": 0.21827103229950592 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2134222515669535e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417052979682934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01479287833894446, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.017091722238510847 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06508360803127289, + "mean_feedback_forward_cosine": 0.9995007384907115, + "mean_feedback_forward_relative_error": 0.030128295220096003, + "min_feedback_forward_cosine": 0.997896134853363 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012462803996308388, + "instruction_rms": 0.0012462803994368237, + "raw_apical_rms": 0.2183206979946904, + "teaching_rms": 0.0012462803996308388, + "traffic_rms": 0.21831621035665064 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2036630891998357e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54250592614894e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014468288615767858, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.01779225636407733 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06507200002670288, + "mean_feedback_forward_cosine": 0.9995008685372092, + "mean_feedback_forward_relative_error": 0.030123845623298125, + "min_feedback_forward_cosine": 0.9978968501091003 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011731846806301597, + "instruction_rms": 0.0011731846806214035, + "raw_apical_rms": 0.2182940879311527, + "teaching_rms": 0.0011731846806301597, + "traffic_rms": 0.21828995860852968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8610290908327443e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542653522729808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014953711034936291, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.015649954566955566 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06506045907735825, + "mean_feedback_forward_cosine": 0.9995010094209151, + "mean_feedback_forward_relative_error": 0.030119523050432855, + "min_feedback_forward_cosine": 0.9978975057601929 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011245090974478294, + "instruction_rms": 0.0011245090976646152, + "raw_apical_rms": 0.21824501696985352, + "teaching_rms": 0.0011245090974478294, + "traffic_rms": 0.21824108245007703 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.377214005060523e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53412665333375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014547173323168636, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.01506383736729622 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06504973769187927, + "mean_feedback_forward_cosine": 0.9995011307976462, + "mean_feedback_forward_relative_error": 0.030115457793528384, + "min_feedback_forward_cosine": 0.9978983402252197 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001117483099401405, + "instruction_rms": 0.0011174830996226938, + "raw_apical_rms": 0.21820990401966323, + "teaching_rms": 0.001117483099401405, + "traffic_rms": 0.21820604765902654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.3820601907355012e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5445933380405018e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014562055954093793, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.014547669793367387 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06503810733556747, + "mean_feedback_forward_cosine": 0.9995012803511186, + "mean_feedback_forward_relative_error": 0.030111161560158837, + "min_feedback_forward_cosine": 0.9978991746902466 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011028086642050866, + "instruction_rms": 0.0011028086642971904, + "raw_apical_rms": 0.21823440846862915, + "teaching_rms": 0.0011028086642050866, + "traffic_rms": 0.21823070044917675 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6005360237159039e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545526299871005e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014770407143025974, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.013574749346300959 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06502733379602432, + "mean_feedback_forward_cosine": 0.9995014103976163, + "mean_feedback_forward_relative_error": 0.03010684595527974, + "min_feedback_forward_cosine": 0.9978997111320496 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010608975832059045, + "instruction_rms": 0.0010608975833145714, + "raw_apical_rms": 0.21813015736353003, + "teaching_rms": 0.0010608975832059045, + "traffic_rms": 0.21812644569294434 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 2.0048213400514214e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.540117475480919e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014574979079508941, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.012621158649325372 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06501596421003342, + "mean_feedback_forward_cosine": 0.9995015556162054, + "mean_feedback_forward_relative_error": 0.03010259923440489, + "min_feedback_forward_cosine": 0.997900664806366 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010572479118043693, + "instruction_rms": 0.0010572479114462404, + "raw_apical_rms": 0.21804978469116781, + "teaching_rms": 0.0010572479118043693, + "traffic_rms": 0.21804615116387266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.698988540965729e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539464958097651e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014874585703551116, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.012497478622347116 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0650041475892067, + "mean_feedback_forward_cosine": 0.9995017008347945, + "mean_feedback_forward_relative_error": 0.0300981184552339, + "min_feedback_forward_cosine": 0.9979014992713928 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010594853302755176, + "instruction_rms": 0.0010594853301246813, + "raw_apical_rms": 0.21809186614572823, + "teaching_rms": 0.0010594853302755176, + "traffic_rms": 0.21808801380600237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.5074134296355624e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5425249564948097e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014690587318314356, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.011729662252962588 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06499096751213074, + "mean_feedback_forward_cosine": 0.9995018623091958, + "mean_feedback_forward_relative_error": 0.03009332650425759, + "min_feedback_forward_cosine": 0.9979022741317749 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011061111347452474, + "instruction_rms": 0.0011061111346037022, + "raw_apical_rms": 0.21795133197846742, + "teaching_rms": 0.0011061111347452474, + "traffic_rms": 0.21794722117908785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.386146095683216e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5434021302429684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014588430622210068, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.012031505008935928 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06497938930988312, + "mean_feedback_forward_cosine": 0.9995020064440641, + "mean_feedback_forward_relative_error": 0.030088558344339784, + "min_feedback_forward_cosine": 0.9979031085968018 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011037157126004146, + "instruction_rms": 0.0011037157127527927, + "raw_apical_rms": 0.2178958414191844, + "teaching_rms": 0.0011037157126004146, + "traffic_rms": 0.21789167785037192 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.5824349475224153e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.540303162050113e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015217674617302382, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.011553875677585602 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06496911495923996, + "mean_feedback_forward_cosine": 0.9995021473277699, + "mean_feedback_forward_relative_error": 0.030084425685080615, + "min_feedback_forward_cosine": 0.9979037642478943 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009879703271231121, + "instruction_rms": 0.0009879703271232561, + "raw_apical_rms": 0.21796217363670467, + "teaching_rms": 0.0009879703271231121, + "traffic_rms": 0.217958711515757 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6918848189106939e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5400295112699034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014525071666045379, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.010497777438797058 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06495882570743561, + "mean_feedback_forward_cosine": 0.9995022687045011, + "mean_feedback_forward_relative_error": 0.030080817749893123, + "min_feedback_forward_cosine": 0.9979043006896973 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008932018895710781, + "instruction_rms": 0.0008932018893837475, + "raw_apical_rms": 0.21782939636326107, + "teaching_rms": 0.0008932018895710781, + "traffic_rms": 0.21782660220360178 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.952057999687895e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5413661997747263e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014655643989231092, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.009360592414438724 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06494715809822083, + "mean_feedback_forward_cosine": 0.9995024074207652, + "mean_feedback_forward_relative_error": 0.030076542632146316, + "min_feedback_forward_cosine": 0.9979052543640137 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009969356816909524, + "instruction_rms": 0.0009969356815726456, + "raw_apical_rms": 0.21782868445685458, + "teaching_rms": 0.0009969356816909524, + "traffic_rms": 0.2178252071654722 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2044817765399785e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5467805004620668e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014545785667416388, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.010036319532692432 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06493602693080902, + "mean_feedback_forward_cosine": 0.9995025493881919, + "mean_feedback_forward_relative_error": 0.03007238908586177, + "min_feedback_forward_cosine": 0.9979058504104614 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009305151011826696, + "instruction_rms": 0.00093051510124445, + "raw_apical_rms": 0.21796102632888387, + "teaching_rms": 0.0009305151011826696, + "traffic_rms": 0.2179578498441756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.516882548457943e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539257472304381e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014519270420902647, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.009270294933617114 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06492490321397781, + "mean_feedback_forward_cosine": 0.9995026902718978, + "mean_feedback_forward_relative_error": 0.030068191784349354, + "min_feedback_forward_cosine": 0.9979067444801331 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009654793644534206, + "instruction_rms": 0.0009654793644367571, + "raw_apical_rms": 0.21787396362379663, + "teaching_rms": 0.0009654793644534206, + "traffic_rms": 0.21787059085325255 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.076040163638936e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543475357971756e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014545164295574387, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.008770944955796004 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06491183489561081, + "mean_feedback_forward_cosine": 0.9995028571649032, + "mean_feedback_forward_relative_error": 0.030063054896891116, + "min_feedback_forward_cosine": 0.9979074001312256 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010621191504510618, + "instruction_rms": 0.0010621191507124818, + "raw_apical_rms": 0.21778792155429455, + "teaching_rms": 0.0010621191504510618, + "traffic_rms": 0.21778389448017127 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2940104454628454e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.546036307815299e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01467553336415156, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.010494573308974504 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0648997575044632, + "mean_feedback_forward_cosine": 0.9995029969648881, + "mean_feedback_forward_relative_error": 0.030058729673989794, + "min_feedback_forward_cosine": 0.9979082345962524 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009133243918846598, + "instruction_rms": 0.0009133243918853063, + "raw_apical_rms": 0.21782069653104164, + "teaching_rms": 0.0009133243918846598, + "traffic_rms": 0.21781752693583212 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6903076982544007e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5568671567537036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014436844543248048, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.00897769759300165 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06488971412181854, + "mean_feedback_forward_cosine": 0.999503137848594, + "mean_feedback_forward_relative_error": 0.03005455881357193, + "min_feedback_forward_cosine": 0.9979089498519897 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009242204102602861, + "instruction_rms": 0.0009242204101807675, + "raw_apical_rms": 0.21779116629990677, + "teaching_rms": 0.0009242204102602861, + "traffic_rms": 0.21778790313516563 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.6618898257547698e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5426019347950658e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01455423309980087, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.008237404524311423 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06487832218408585, + "mean_feedback_forward_cosine": 0.9995032787322998, + "mean_feedback_forward_relative_error": 0.030050031688403, + "min_feedback_forward_cosine": 0.9979096055030823 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009365221833486327, + "instruction_rms": 0.0009365221832434332, + "raw_apical_rms": 0.21761712505623068, + "teaching_rms": 0.0009365221833486327, + "traffic_rms": 0.21761389262363148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4782013124658988e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5337201187037823e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014548863837105585, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.008197247269898653 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06486883759498596, + "mean_feedback_forward_cosine": 0.9995033936067061, + "mean_feedback_forward_relative_error": 0.03004643761298873, + "min_feedback_forward_cosine": 0.9979102611541748 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008333562779346669, + "instruction_rms": 0.0008333562778890741, + "raw_apical_rms": 0.21760150552138213, + "teaching_rms": 0.0008333562779346669, + "traffic_rms": 0.21759892968281624 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.3724050518248987e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5392741898444937e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01443405107828143, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.0069121859993040565 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.064858078956604, + "mean_feedback_forward_cosine": 0.9995035388252952, + "mean_feedback_forward_relative_error": 0.030042262527753005, + "min_feedback_forward_cosine": 0.9979109168052673 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008470358747454884, + "instruction_rms": 0.000847035874848523, + "raw_apical_rms": 0.2177094235508519, + "teaching_rms": 0.0008470358747454884, + "traffic_rms": 0.21770660806724032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.1853709480647012e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5463040858939075e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014599937611635421, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.007263571571558714 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06484507769346237, + "mean_feedback_forward_cosine": 0.9995036992159757, + "mean_feedback_forward_relative_error": 0.030037531028078362, + "min_feedback_forward_cosine": 0.997911810874939 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009227820317200313, + "instruction_rms": 0.0009227820313578128, + "raw_apical_rms": 0.21778683776085886, + "teaching_rms": 0.0009227820317200313, + "traffic_rms": 0.21778349215921253 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.518674679346077e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541176001662854e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014500019283860853, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.008067312535867096 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06483304500579834, + "mean_feedback_forward_cosine": 0.9995038520206104, + "mean_feedback_forward_relative_error": 0.03003282580016689, + "min_feedback_forward_cosine": 0.9979126453399658 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000904557197257956, + "instruction_rms": 0.0009045571968954518, + "raw_apical_rms": 0.21781838418552885, + "teaching_rms": 0.000904557197257956, + "traffic_rms": 0.21781511411048748 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2415722494552028e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5446335777436418e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01467407361531207, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008004105790928006 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06482066214084625, + "mean_feedback_forward_cosine": 0.9995040199973366, + "mean_feedback_forward_relative_error": 0.03002789347347888, + "min_feedback_forward_cosine": 0.9979134798049927 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008853743437169138, + "instruction_rms": 0.0008853743434541338, + "raw_apical_rms": 0.21778107415348422, + "teaching_rms": 0.0008853743437169138, + "traffic_rms": 0.21777786626930937 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4368413303600391e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544887959792478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015137492293296017, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.0077969551001489164 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06480864435434341, + "mean_feedback_forward_cosine": 0.9995041597973217, + "mean_feedback_forward_relative_error": 0.03002361799166961, + "min_feedback_forward_cosine": 0.9979141354560852 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008391028517926895, + "instruction_rms": 0.000839102852031073, + "raw_apical_rms": 0.21773445581837678, + "teaching_rms": 0.0008391028517926895, + "traffic_rms": 0.2177316099056005 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4985285190505238e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5370226743325675e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014714404022450592, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.006946576703712344 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0647956058382988, + "mean_feedback_forward_cosine": 0.9995043104345148, + "mean_feedback_forward_relative_error": 0.030019104810939595, + "min_feedback_forward_cosine": 0.9979149699211121 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008154932633886795, + "instruction_rms": 0.0008154932633458998, + "raw_apical_rms": 0.21768122263750744, + "teaching_rms": 0.0008154932633886795, + "traffic_rms": 0.21767842680204852 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.458872702642111e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5427034575513562e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01459319958469563, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.007164511053469032 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06478343158960342, + "mean_feedback_forward_cosine": 0.999504443732175, + "mean_feedback_forward_relative_error": 0.030014908881011333, + "min_feedback_forward_cosine": 0.9979156851768494 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008122454089370418, + "instruction_rms": 0.0008122454089585955, + "raw_apical_rms": 0.21757522233357648, + "teaching_rms": 0.0008122454089370418, + "traffic_rms": 0.21757248699384205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 1.9975442455688608e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5396378177493007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014840558629132505, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.006613630339205265 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06477167457342148, + "mean_feedback_forward_cosine": 0.9995045867833224, + "mean_feedback_forward_relative_error": 0.030010619895024734, + "min_feedback_forward_cosine": 0.9979163408279419 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000816638778300246, + "instruction_rms": 0.0008166387786254303, + "raw_apical_rms": 0.21757390088774958, + "teaching_rms": 0.000816638778300246, + "traffic_rms": 0.2175711414491402 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4611411813802988e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53968035951542e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014801631390164239, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.006615142454504967 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06476040184497833, + "mean_feedback_forward_cosine": 0.9995047298344699, + "mean_feedback_forward_relative_error": 0.030006251331757414, + "min_feedback_forward_cosine": 0.997917115688324 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008299731161878149, + "instruction_rms": 0.0008299731164876084, + "raw_apical_rms": 0.21753630208598787, + "teaching_rms": 0.0008299731161878149, + "traffic_rms": 0.21753350667161842 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4189680541676353e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.546554309558204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014689996827664377, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.006605607134364545 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06474673002958298, + "mean_feedback_forward_cosine": 0.9995049259879373, + "mean_feedback_forward_relative_error": 0.03000120551071384, + "min_feedback_forward_cosine": 0.9979180097579956 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008570837406486209, + "instruction_rms": 0.0008570837405911377, + "raw_apical_rms": 0.21749114617243778, + "teaching_rms": 0.0008570837406486209, + "traffic_rms": 0.21748807498193576 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2037280328058841e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542142572168379e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014435914248308288, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.006848038278520107 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06473147124052048, + "mean_feedback_forward_cosine": 0.9995050863786177, + "mean_feedback_forward_relative_error": 0.029996219649910928, + "min_feedback_forward_cosine": 0.9979190826416016 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008380833239710877, + "instruction_rms": 0.0008380833240598763, + "raw_apical_rms": 0.21745341050619166, + "teaching_rms": 0.0008380833239710877, + "traffic_rms": 0.21745031588926456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.735748753617372e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54118015201925e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014786675804777332, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.006193644772060215 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06471840292215347, + "mean_feedback_forward_cosine": 0.9995052576065063, + "mean_feedback_forward_relative_error": 0.029991317582740026, + "min_feedback_forward_cosine": 0.9979197978973389 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008256146901548177, + "instruction_rms": 0.0008256146902481165, + "raw_apical_rms": 0.2174068893619909, + "teaching_rms": 0.0008256146901548177, + "traffic_rms": 0.21740409180375106 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.326285143932182e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5454300474418304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014689367005480774, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.006454593700170517 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0647064596414566, + "mean_feedback_forward_cosine": 0.9995054006576538, + "mean_feedback_forward_relative_error": 0.029986666557802397, + "min_feedback_forward_cosine": 0.997920572757721 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007791413530249653, + "instruction_rms": 0.0007791413530333879, + "raw_apical_rms": 0.2173478491451276, + "teaching_rms": 0.0007791413530249653, + "traffic_rms": 0.21734518388887725 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.6995947760278796e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5411865634509273e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014772910919759207, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006066935067735612 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06469673663377762, + "mean_feedback_forward_cosine": 0.9995055263692683, + "mean_feedback_forward_relative_error": 0.02998274332759055, + "min_feedback_forward_cosine": 0.997921347618103 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007483332100994911, + "instruction_rms": 0.000748333209860165, + "raw_apical_rms": 0.21736623109221268, + "teaching_rms": 0.0007483332100994911, + "traffic_rms": 0.21736379379845625 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.7301458399288094e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5446078231496358e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01485945847067322, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.005622729334756732 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646849051117897, + "mean_feedback_forward_cosine": 0.999505679173903, + "mean_feedback_forward_relative_error": 0.029978170212019574, + "min_feedback_forward_cosine": 0.9979221224784851 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007666970909328611, + "instruction_rms": 0.0007666970909207206, + "raw_apical_rms": 0.21735056299819014, + "teaching_rms": 0.0007666970909328611, + "traffic_rms": 0.21734800189541878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.5576570433495363e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.540577537393375e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014714248909018339, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.005723501109331846 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06467004120349884, + "mean_feedback_forward_cosine": 0.9995058612389998, + "mean_feedback_forward_relative_error": 0.02997291199862957, + "min_feedback_forward_cosine": 0.9979228973388672 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007845029950275426, + "instruction_rms": 0.0007845029947168352, + "raw_apical_rms": 0.21738187409485815, + "teaching_rms": 0.0007845029950275426, + "traffic_rms": 0.2173788851185387 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4064256959045451e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5367123453454566e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014728014148771583, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.006031060835011304 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465739756822586, + "mean_feedback_forward_cosine": 0.9995060162110763, + "mean_feedback_forward_relative_error": 0.029968157766217536, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000778150866247698, + "instruction_rms": 0.0007781508661009998, + "raw_apical_rms": 0.2173779227603393, + "teaching_rms": 0.000778150866247698, + "traffic_rms": 0.21737513508900885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.697241458487042e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.549316063606117e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01487251115746788, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005693117426261306 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465724855661392, + "mean_feedback_forward_cosine": 0.999506023797122, + "mean_feedback_forward_relative_error": 0.029968085732649672, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007336816546237581, + "instruction_rms": 0.0007336816545011432, + "raw_apical_rms": 0.217371156494388, + "teaching_rms": 0.0007336816546237581, + "traffic_rms": 0.21736897742453662 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 1.8313436100925173e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5456454865421184e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014622755278931707, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.0052648243070393804 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465716660022736, + "mean_feedback_forward_cosine": 0.999506017294797, + "mean_feedback_forward_relative_error": 0.02996802329170433, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006499877582761629, + "instruction_rms": 0.0006499877582179854, + "raw_apical_rms": 0.21738199483031725, + "teaching_rms": 0.0006499877582761629, + "traffic_rms": 0.21737998388546803 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.4211714471912273e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5390023372203127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015088620538595623, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004518425296097994 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646570697426796, + "mean_feedback_forward_cosine": 0.999506017294797, + "mean_feedback_forward_relative_error": 0.029967964508316732, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006424708529576893, + "instruction_rms": 0.0006424708526524166, + "raw_apical_rms": 0.21738588404642106, + "teaching_rms": 0.0006424708529576893, + "traffic_rms": 0.21738394391061916 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6667971181050234e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53441945221952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014840298852637109, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.004693933547213674 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465697288513184, + "mean_feedback_forward_cosine": 0.9995060302994468, + "mean_feedback_forward_relative_error": 0.02996790644458749, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006355975641246289, + "instruction_rms": 0.0006355975638217237, + "raw_apical_rms": 0.21737219777478292, + "teaching_rms": 0.0006355975641246289, + "traffic_rms": 0.21737017795801059 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7290426512772683e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5412156285035388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014799775741982742, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004402383487150073 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465689092874527, + "mean_feedback_forward_cosine": 0.9995060151273554, + "mean_feedback_forward_relative_error": 0.029967856051569635, + "min_feedback_forward_cosine": 0.9979236721992493 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005838366616606998, + "instruction_rms": 0.0005838366617374863, + "raw_apical_rms": 0.21738252372913064, + "teaching_rms": 0.0005838366616606998, + "traffic_rms": 0.21738087089018013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.411203243151249e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5369329955990666e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014650811609760516, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004063497681729496 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646568089723587, + "mean_feedback_forward_cosine": 0.9995060151273554, + "mean_feedback_forward_relative_error": 0.029967801010405474, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005977112082014249, + "instruction_rms": 0.0005977112077572961, + "raw_apical_rms": 0.2173785091480081, + "teaching_rms": 0.0005977112082014249, + "traffic_rms": 0.21737685000147602 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.2497670809352712e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5394922282883387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014765513541217284, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004217370530590415 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465671211481094, + "mean_feedback_forward_cosine": 0.9995060216296803, + "mean_feedback_forward_relative_error": 0.029967742472548378, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006564801779574908, + "instruction_rms": 0.0006564801779524528, + "raw_apical_rms": 0.2173975074946715, + "teaching_rms": 0.0006564801779574908, + "traffic_rms": 0.21739536422206815 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.7395591100921592e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5423988215072534e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014798289134398644, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004837606736086309 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465663760900497, + "mean_feedback_forward_cosine": 0.999506017294797, + "mean_feedback_forward_relative_error": 0.029967696194282988, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005231504160459126, + "instruction_rms": 0.0005231504161347653, + "raw_apical_rms": 0.21740460564968178, + "teaching_rms": 0.0005231504160459126, + "traffic_rms": 0.21740330273734604 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.0684335904898035e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5408948494657622e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014793308600444385, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.0036588855369761586 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.064656563103199, + "mean_feedback_forward_cosine": 0.9995060443878174, + "mean_feedback_forward_relative_error": 0.029967649213292383, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005640463211356523, + "instruction_rms": 0.0005640463211281321, + "raw_apical_rms": 0.21741686951987077, + "teaching_rms": 0.0005640463211356523, + "traffic_rms": 0.21741523138095503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.814800185684362e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5387875906848967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014930301889274982, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.003823152501806617 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465646624565125, + "mean_feedback_forward_cosine": 0.9995060346343301, + "mean_feedback_forward_relative_error": 0.029967594299126755, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006005586770164504, + "instruction_rms": 0.0006005586768464487, + "raw_apical_rms": 0.21741327747622716, + "teaching_rms": 0.0006005586770164504, + "traffic_rms": 0.21741137384985465 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.6142958259024454e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5387198470691356e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014739232013794484, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.004214129344187677 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465637683868408, + "mean_feedback_forward_cosine": 0.9995060270482844, + "mean_feedback_forward_relative_error": 0.029967545997351408, + "min_feedback_forward_cosine": 0.997923731803894 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005662638416434736, + "instruction_rms": 0.0005662638418921005, + "raw_apical_rms": 0.21741102345080152, + "teaching_rms": 0.0005662638416434736, + "traffic_rms": 0.21740931423612808 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.445411470513136e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5353476278442575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014805875585444935, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0038237521389871834 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465629488229752, + "mean_feedback_forward_cosine": 0.9995060270482844, + "mean_feedback_forward_relative_error": 0.029967497069049966, + "min_feedback_forward_cosine": 0.997923731803894 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005777983687812732, + "instruction_rms": 0.0005777983689998404, + "raw_apical_rms": 0.21739999761080617, + "teaching_rms": 0.0005777983687812732, + "traffic_rms": 0.21739840666902674 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.7733233903527434e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5407223240619934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014928091566826179, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.004213563338406384 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465622782707214, + "mean_feedback_forward_cosine": 0.9995060378854925, + "mean_feedback_forward_relative_error": 0.029967449080537667, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005344018490385648, + "instruction_rms": 0.0005344018490286388, + "raw_apical_rms": 0.2173903266768718, + "teaching_rms": 0.0005344018490385648, + "traffic_rms": 0.21738883866752137 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.8528919554228196e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5370226856621487e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01478576908927508, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.0038587893366254867 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465614587068558, + "mean_feedback_forward_cosine": 0.9995060270482844, + "mean_feedback_forward_relative_error": 0.02996739289638671, + "min_feedback_forward_cosine": 0.997923731803894 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00058741375419832, + "instruction_rms": 0.0005874137543652931, + "raw_apical_rms": 0.2174028632316789, + "teaching_rms": 0.00058741375419832, + "traffic_rms": 0.21740090854154104 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 2.3538157734037668e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5385025181287067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014802609587018959, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.00428344938647002 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465606391429901, + "mean_feedback_forward_cosine": 0.999506029215726, + "mean_feedback_forward_relative_error": 0.02996734015812928, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000586293719120961, + "instruction_rms": 0.0005862937191397492, + "raw_apical_rms": 0.2174110437708709, + "teaching_rms": 0.000586293719120961, + "traffic_rms": 0.21740937994403361 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.537450202704349e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538488882263741e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014767745614436181, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.004168851910606027 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465597450733185, + "mean_feedback_forward_cosine": 0.9995060498064214, + "mean_feedback_forward_relative_error": 0.029967289621179755, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005711286295311961, + "instruction_rms": 0.0005711286296726123, + "raw_apical_rms": 0.21742013608232025, + "teaching_rms": 0.0005711286295311961, + "traffic_rms": 0.21741843722367432 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.7039680244579358e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53934933282661e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014841632680235542, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003980741167739034 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465589255094528, + "mean_feedback_forward_cosine": 0.9995060368017716, + "mean_feedback_forward_relative_error": 0.029967241573401474, + "min_feedback_forward_cosine": 0.997923731803894 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005488385329512543, + "instruction_rms": 0.0005488385327596644, + "raw_apical_rms": 0.21742746812759606, + "teaching_rms": 0.0005488385329512543, + "traffic_rms": 0.21742596438432119 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.6996645513245312e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537199711002185e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014674170605712869, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.003744773309454322 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646558329463005, + "mean_feedback_forward_cosine": 0.9995060400529341, + "mean_feedback_forward_relative_error": 0.02996719598939473, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005337192166041582, + "instruction_rms": 0.0005337192165815987, + "raw_apical_rms": 0.21742422693084587, + "teaching_rms": 0.0005337192166041582, + "traffic_rms": 0.2174228491767559 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.7608694236059819e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5406195897977895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014804681210761178, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.003526218791976571 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465577334165573, + "mean_feedback_forward_cosine": 0.9995060335506093, + "mean_feedback_forward_relative_error": 0.029967155121266843, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004996069556226524, + "instruction_rms": 0.0004996069556369657, + "raw_apical_rms": 0.21741222278816413, + "teaching_rms": 0.0004996069556226524, + "traffic_rms": 0.21741099645953593 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 2.0968676039956335e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539053017426234e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015146897537673782, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.003425298108793795 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465571373701096, + "mean_feedback_forward_cosine": 0.9995060389692133, + "mean_feedback_forward_relative_error": 0.02996711063791405, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005247809982073545, + "instruction_rms": 0.0005247809980287322, + "raw_apical_rms": 0.21740645757151203, + "teaching_rms": 0.0005247809982073545, + "traffic_rms": 0.21740500772533136 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.419543735892148e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544157580582638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01471563105188694, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0035274609739892183 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465563923120499, + "mean_feedback_forward_cosine": 0.999506046555259, + "mean_feedback_forward_relative_error": 0.029967066433958033, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005108157900904676, + "instruction_rms": 0.0005108157901464399, + "raw_apical_rms": 0.2174094285213894, + "teaching_rms": 0.0005108157900904676, + "traffic_rms": 0.2174081244394395 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4977540274685452e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5412310028184802e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015139901016237727, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.003486002595461905 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465555727481842, + "mean_feedback_forward_cosine": 0.9995060443878174, + "mean_feedback_forward_relative_error": 0.02996701556681232, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005845881125955994, + "instruction_rms": 0.00058458811230708, + "raw_apical_rms": 0.2173959602032729, + "teaching_rms": 0.0005845881125955994, + "traffic_rms": 0.21739426045309143 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4391186198281503e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363599367415803e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014783370626484523, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003774541051406413 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465548276901245, + "mean_feedback_forward_cosine": 0.9995060443878174, + "mean_feedback_forward_relative_error": 0.029966969788074494, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005172267136059235, + "instruction_rms": 0.0005172267134339713, + "raw_apical_rms": 0.21739570198721703, + "teaching_rms": 0.0005172267136059235, + "traffic_rms": 0.21739415631778516 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.7663832352354802e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53763648735008e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015075271571038825, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0034943548517813905 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465540826320648, + "mean_feedback_forward_cosine": 0.9995060346343301, + "mean_feedback_forward_relative_error": 0.029966924246400593, + "min_feedback_forward_cosine": 0.9979237914085388 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047680270575156967, + "instruction_rms": 0.0004768027057745421, + "raw_apical_rms": 0.21740465935505576, + "teaching_rms": 0.00047680270575156967, + "traffic_rms": 0.2174032656897058 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894432544708, + "maximum_absolute_post_projection_soma_slope": 1.6243130573911913e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5410672530972694e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014770830971317052, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0033188046364113688 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646553561091423, + "mean_feedback_forward_cosine": 0.999506046555259, + "mean_feedback_forward_relative_error": 0.0299668820744211, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004823949782915145, + "instruction_rms": 0.0004823949781897861, + "raw_apical_rms": 0.2174034631546331, + "teaching_rms": 0.0004823949782915145, + "traffic_rms": 0.21740222987061142 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.3032894730567932, + "maximum_absolute_post_projection_soma_slope": 2.0330870142970525e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539493085993386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014719714613152765, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.003235620997995138 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465528905391693, + "mean_feedback_forward_cosine": 0.9995060476389799, + "mean_feedback_forward_relative_error": 0.029966836956075646, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005192002842949462, + "instruction_rms": 0.0005192002841667181, + "raw_apical_rms": 0.21740963727154622, + "teaching_rms": 0.0005192002842949462, + "traffic_rms": 0.21740824891497282 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4374454337939824e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54143562147924e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015180627090760313, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0036198783402796833 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465522944927216, + "mean_feedback_forward_cosine": 0.9995060378854925, + "mean_feedback_forward_relative_error": 0.02996679348871112, + "min_feedback_forward_cosine": 0.9979238510131836 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047765511720424756, + "instruction_rms": 0.0004776551170704759, + "raw_apical_rms": 0.21740837779312225, + "teaching_rms": 0.00047765511720424756, + "traffic_rms": 0.2174072304949405 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.6102829647479666e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5392311416945337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014867588664745803, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0031951745545119046 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465516984462738, + "mean_feedback_forward_cosine": 0.9995060563087463, + "mean_feedback_forward_relative_error": 0.02996675068173896, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005090372178668618, + "instruction_rms": 0.000509037217898844, + "raw_apical_rms": 0.21742121981918439, + "teaching_rms": 0.0005090372178668618, + "traffic_rms": 0.21741990100883468 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7437180588331103e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545058390190179e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014876043814117336, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.003262077833302319 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465507298707962, + "mean_feedback_forward_cosine": 0.9995060433040965, + "mean_feedback_forward_relative_error": 0.02996670018712228, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000569124005517209, + "instruction_rms": 0.0005691240056004799, + "raw_apical_rms": 0.2174115342567077, + "teaching_rms": 0.000569124005517209, + "traffic_rms": 0.21740992758003677 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.7676303798452864e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534766744941932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014864871087457533, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0038019757165014744 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465500593185425, + "mean_feedback_forward_cosine": 0.9995060454715382, + "mean_feedback_forward_relative_error": 0.029966658396138386, + "min_feedback_forward_cosine": 0.997923731803894 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048721177301676153, + "instruction_rms": 0.00048721177276412584, + "raw_apical_rms": 0.21739991392401944, + "teaching_rms": 0.00048721177301676153, + "traffic_rms": 0.21739872380390068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.559779718718346e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5451993017862373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014874742900861508, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0032574594100378456 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465493887662888, + "mean_feedback_forward_cosine": 0.9995060454715382, + "mean_feedback_forward_relative_error": 0.02996661550450054, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000494028143340967, + "instruction_rms": 0.0004940281432991926, + "raw_apical_rms": 0.21739441667603263, + "teaching_rms": 0.000494028143340967, + "traffic_rms": 0.21739315922411628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4297611983238312e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5428185038179038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014802584077321858, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.003330545827448368 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646548792719841, + "mean_feedback_forward_cosine": 0.9995060487227007, + "mean_feedback_forward_relative_error": 0.02996657093648206, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005206205105650402, + "instruction_rms": 0.0005206205107205834, + "raw_apical_rms": 0.21739966314334672, + "teaching_rms": 0.0005206205105650402, + "traffic_rms": 0.2173982548642933 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.1608278782659909e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544242047799201e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014724816158551066, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0032657277070358396 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465479731559753, + "mean_feedback_forward_cosine": 0.9995060454715382, + "mean_feedback_forward_relative_error": 0.029966524768282068, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004978860861837604, + "instruction_rms": 0.0004978860863518811, + "raw_apical_rms": 0.21740888659840904, + "teaching_rms": 0.0004978860861837604, + "traffic_rms": 0.21740743908307528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.8766252196655842e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5432152921863694e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014810028981526032, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0034700161396712063 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465473026037216, + "mean_feedback_forward_cosine": 0.9995060693133961, + "mean_feedback_forward_relative_error": 0.02996648144484921, + "min_feedback_forward_cosine": 0.9979240298271179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005215601427734619, + "instruction_rms": 0.0005215601425838428, + "raw_apical_rms": 0.2174011236627009, + "teaching_rms": 0.0005215601427734619, + "traffic_rms": 0.21739975292772665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.223177434894751e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542862469549623e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015025801284942993, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0033588976715505123 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465465575456619, + "mean_feedback_forward_cosine": 0.9995060519738631, + "mean_feedback_forward_relative_error": 0.029966438849541276, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000499381375764398, + "instruction_rms": 0.0004993813756396742, + "raw_apical_rms": 0.21740655922939264, + "teaching_rms": 0.000499381375764398, + "traffic_rms": 0.21740526230389107 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.4619259047776723e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5409276797362904e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014788344967447473, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0033693819453567268 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465459614992142, + "mean_feedback_forward_cosine": 0.9995060508901422, + "mean_feedback_forward_relative_error": 0.02996639484878291, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004621170943048924, + "instruction_rms": 0.00046211709406335363, + "raw_apical_rms": 0.21741432584785536, + "teaching_rms": 0.0004621170943048924, + "traffic_rms": 0.2174130967605341 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.376182012791105e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539613960120708e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014716894426569689, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0031655293280258776 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465452909469604, + "mean_feedback_forward_cosine": 0.9995060552250256, + "mean_feedback_forward_relative_error": 0.029966348680582913, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005219874691613879, + "instruction_rms": 0.0005219874689990842, + "raw_apical_rms": 0.21740740280217968, + "teaching_rms": 0.0005219874691613879, + "traffic_rms": 0.2174059162679814 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.613528866073466e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539200203000518e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014772549855444916, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.00337699354454875 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465446203947067, + "mean_feedback_forward_cosine": 0.9995060552250256, + "mean_feedback_forward_relative_error": 0.02996630777858875, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004401869693800428, + "instruction_rms": 0.00044018696941048116, + "raw_apical_rms": 0.21739349082914738, + "teaching_rms": 0.0004401869693800428, + "traffic_rms": 0.21739242801023695 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328935384750366, + "maximum_absolute_post_projection_soma_slope": 1.3805274079459195e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.54102377627672e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014765603761886445, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.0030622552121989427 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646543800830841, + "mean_feedback_forward_cosine": 0.9995060498064214, + "mean_feedback_forward_relative_error": 0.029966263608499008, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005126393015931346, + "instruction_rms": 0.0005126393016675731, + "raw_apical_rms": 0.21738862227333874, + "teaching_rms": 0.0005126393015931346, + "traffic_rms": 0.21738720255921504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.492823713533653e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.542992685895887e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014703011810006272, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003446732765249908 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465432047843933, + "mean_feedback_forward_cosine": 0.9995060736482794, + "mean_feedback_forward_relative_error": 0.029966227202252907, + "min_feedback_forward_cosine": 0.9979240298271179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042896103254150096, + "instruction_rms": 0.00042896103243049427, + "raw_apical_rms": 0.21738216678016126, + "teaching_rms": 0.00042896103254150096, + "traffic_rms": 0.2173811609021421 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.416471064885627e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541990399225296e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014830609030900289, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.002722090342491865 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465424597263336, + "mean_feedback_forward_cosine": 0.9995060693133961, + "mean_feedback_forward_relative_error": 0.02996617847714912, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005017619126451125, + "instruction_rms": 0.0005017619125553546, + "raw_apical_rms": 0.21737164417073454, + "teaching_rms": 0.0005017619126451125, + "traffic_rms": 0.21737014588535458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.6515421918938955e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5456002938367865e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014759938623378302, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0033543736807256936 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0646541640162468, + "mean_feedback_forward_cosine": 0.9995060703971169, + "mean_feedback_forward_relative_error": 0.029966128634458238, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005124443840093078, + "instruction_rms": 0.0005124443841018715, + "raw_apical_rms": 0.21738106285126324, + "teaching_rms": 0.0005124443840093078, + "traffic_rms": 0.21737949702215628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.374504989826164e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5445737203376625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014916032998438793, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.003268081470038742 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465410441160202, + "mean_feedback_forward_cosine": 0.9995060671459545, + "mean_feedback_forward_relative_error": 0.029966086962006307, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004652341979512973, + "instruction_rms": 0.0004652341979292752, + "raw_apical_rms": 0.21737825899908592, + "teaching_rms": 0.0004652341979512973, + "traffic_rms": 0.21737693815033785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.3245708885278873e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.546000521723271e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015385667096859386, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0031492233463004233 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465402990579605, + "mean_feedback_forward_cosine": 0.9995060573924671, + "mean_feedback_forward_relative_error": 0.02996604374017228, + "min_feedback_forward_cosine": 0.9979239106178284 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047173044018590624, + "instruction_rms": 0.0004717304399300386, + "raw_apical_rms": 0.21736509679969196, + "teaching_rms": 0.00047173044018590624, + "traffic_rms": 0.21736377003916488 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.351129270688034e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5388899873359245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014724101628322498, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0031086916375532747 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465397775173187, + "mean_feedback_forward_cosine": 0.9995060714808377, + "mean_feedback_forward_relative_error": 0.029966003667901864, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004699245203918985, + "instruction_rms": 0.0004699245202037794, + "raw_apical_rms": 0.21737071026339022, + "teaching_rms": 0.0004699245203918985, + "traffic_rms": 0.21736946554807926 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.4053436814265297e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5402495016101293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01471504414380081, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.003135574822369963 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465393304824829, + "mean_feedback_forward_cosine": 0.9995060790668834, + "mean_feedback_forward_relative_error": 0.02996596359563145, + "min_feedback_forward_cosine": 0.9979240298271179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00043156756303708635, + "instruction_rms": 0.00043156756290855933, + "raw_apical_rms": 0.21736636893496109, + "teaching_rms": 0.00043156756303708635, + "traffic_rms": 0.21736533672306935 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.328577212689197e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417359676759665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014745400055178466, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0026692149826884268 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465386599302292, + "mean_feedback_forward_cosine": 0.9995060649785128, + "mean_feedback_forward_relative_error": 0.02996592190624638, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004731337698227652, + "instruction_rms": 0.0004731337698627747, + "raw_apical_rms": 0.21737531541875718, + "teaching_rms": 0.0004731337698227652, + "traffic_rms": 0.2173741105770319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 2.0235361830600596e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5428865423971323e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014727790364439046, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.002996583756618202 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465378403663635, + "mean_feedback_forward_cosine": 0.9995060736482794, + "mean_feedback_forward_relative_error": 0.029965876627036116, + "min_feedback_forward_cosine": 0.9979240298271179 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004894008205334953, + "instruction_rms": 0.0004894008204920032, + "raw_apical_rms": 0.21737143889129606, + "teaching_rms": 0.0004894008205334953, + "traffic_rms": 0.21737008850784664 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.630112365091918e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.544127677474315e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014804889972159688, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0033445166710577906 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465370953083038, + "mean_feedback_forward_cosine": 0.9995060801506043, + "mean_feedback_forward_relative_error": 0.02996583070436662, + "min_feedback_forward_cosine": 0.9979239702224731 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004714014026243369, + "instruction_rms": 0.00047140140240393275, + "raw_apical_rms": 0.21736337218665547, + "teaching_rms": 0.0004714014026243369, + "traffic_rms": 0.21736210394017905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328938364982605, + "maximum_absolute_post_projection_soma_slope": 1.5625910521066544e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5403167405645478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014824347345283921, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.002965526859872043 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06465364992618561, + "mean_feedback_forward_cosine": 0.9995060747320001, + "mean_feedback_forward_relative_error": 0.029965790149501778, + "min_feedback_forward_cosine": 0.9979240894317627 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00045414121840758485, + "instruction_rms": 0.000454141218399506, + "raw_apical_rms": 0.21735632978428515, + "teaching_rms": 0.00045414121840758485, + "traffic_rms": 0.21735519234462988 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.30328941345214844, + "maximum_absolute_post_projection_soma_slope": 1.567328240525967e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5423811025095633e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014725825154811948, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.0027749313162826 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9264, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.40963050231933595, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3324800000, + "peak_memory_reserved_bytes": 5553258496, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 1.3212634542014712e-07, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 1.3212634542014712e-07, + "min_residual_soma_slope": -1.2691913298112922e-07, + "mse": 2.1058514214185065e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.2224856950565e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.2224856950565e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 12, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.02486515045166, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.06300473213195801, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 26385.03307390213, + "train_wall_s": 26378.69224047661 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.0627080574631691, + 0.03706730529665947, + 0.026097532361745834, + 0.025997934862971306, + 0.015059462748467922, + 0.017114033922553062, + 0.010142219252884388, + 0.012665885500609875, + 0.00798695906996727, + 0.01084459200501442, + 0.0064156330190598965, + 0.00910227745771408, + 0.005154291167855263, + 0.006877919193357229, + 0.004515109118074179, + 0.0063164751045405865, + 0.003920458722859621, + 0.005357971414923668, + 0.003439005697146058, + 0.007975700311362743, + 0.005291788373142481, + 0.006808620411902666, + 0.004765127785503864, + 0.0066861556842923164, + 0.004128842614591122, + 0.0060616303235292435, + 0.003541220910847187, + 0.005087018013000488, + 0.003170633688569069, + 0.00439643207937479, + 0.002961694495752454, + 0.004410099238157272, + 0.002663350198417902, + 0.0037906498182564974, + 0.0024551311507821083, + 0.003720915177837014, + 0.0022440673783421516, + 0.005535278003662825, + 0.003655515145510435, + 0.00500504532828927, + 0.0033655446022748947, + 0.004662506282329559, + 0.003106443677097559, + 0.004185087513178587, + 0.002920798258855939, + 0.003907869104295969, + 0.0027478388510644436, + 0.0037486148066818714, + 0.0026096648070961237, + 0.0033798557706177235, + 0.002526075579226017, + 0.0031582352239638567, + 0.0024816126096993685, + 0.0029130764305591583, + 0.002540357643738389 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 4.000000476837158, + 4.0, + 3.999999761581421, + 3.999999761581421, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.000000476837158, + 3.999999523162842, + 3.999999523162842, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.33750131726264954, + 0.19590385258197784, + 0.08895129710435867, + 0.1366502046585083, + 0.04054412990808487, + 0.0918569266796112, + 0.023009901866316795, + 0.06589116156101227, + 0.015995139256119728, + 0.0597301684319973, + 0.011368107981979847, + 0.05139009281992912, + 0.00864375289529562, + 0.03534276410937309, + 0.007082617841660976, + 0.03329198807477951, + 0.0058189029805362225, + 0.027272045612335205, + 0.004795322194695473, + 0.04221294820308685, + 0.009607328101992607, + 0.0358678475022316, + 0.00799160823225975, + 0.03545752912759781, + 0.006400048732757568, + 0.03432459384202957, + 0.00509628327563405, + 0.027625061571598053, + 0.004334445111453533, + 0.022093990817666054, + 0.003876881441101432, + 0.023369548842310905, + 0.0033206844236701727, + 0.019791830331087112, + 0.0029678642749786377, + 0.019896846264600754, + 0.0026025776751339436, + 0.03057720698416233, + 0.005607756786048412, + 0.0264064222574234, + 0.004839103203266859, + 0.025382149964571, + 0.004209002945572138, + 0.021822823211550713, + 0.0037249738816171885, + 0.02059056982398033, + 0.0033972274977713823, + 0.019883370026946068, + 0.00307767023332417, + 0.01821579970419407, + 0.002869992982596159, + 0.0169074647128582, + 0.0026983791030943394, + 0.015219924971461296, + 0.002645905362442136 + ], + "unscaled_traffic_rms": [ + 0.7432036995887756, + 0.7568469047546387, + 1.1735650300979614, + 0.7610068321228027, + 1.4857354164123535, + 0.7452474236488342, + 1.7631051540374756, + 0.7688974142074585, + 1.9973465204238892, + 0.7262388467788696, + 2.2574145793914795, + 0.7084850072860718, + 2.385209798812866, + 0.7784246206283569, + 2.549966335296631, + 0.7589183449745178, + 2.694981336593628, + 0.7858554720878601, + 2.86863374710083, + 0.7557586431503296, + 2.2032299041748047, + 0.759300708770752, + 2.385065793991089, + 0.7542720437049866, + 2.5805070400238037, + 0.7063891887664795, + 2.779453754425049, + 0.7365801334381104, + 2.92598819732666, + 0.7959507703781128, + 3.0557494163513184, + 0.7548454403877258, + 3.2081942558288574, + 0.7661039233207703, + 3.3089535236358643, + 0.7480412125587463, + 3.4489920139312744, + 0.7241051197052002, + 2.6074705123901367, + 0.7581557631492615, + 2.7819573879241943, + 0.734769344329834, + 2.9521896839141846, + 0.767102837562561, + 3.1364495754241943, + 0.7591570615768433, + 3.2353899478912354, + 0.7541206479072571, + 3.391740560531616, + 0.7421811819076538, + 3.5206713676452637, + 0.7471812963485718, + 3.6786715984344482, + 0.7655954957008362, + 3.8404359817504883 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250441282772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 127796290519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 5010596513857536 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json new file mode 100644 index 0000000..94c1f28 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json @@ -0,0 +1,7293 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 13, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 13, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5013, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9994251562489403, + "feedback_forward_cosine": [ + 0.9995708465576172, + 0.999543309211731, + 0.9994069337844849, + 0.9993953704833984, + 0.9995234608650208, + 0.9994070529937744, + 0.9995660781860352, + 0.999530553817749, + 0.9992876648902893, + 0.999217689037323, + 0.9996702075004578, + 0.9996304512023926, + 0.9995496273040771, + 0.9995319843292236, + 0.9997491240501404, + 0.9996465444564819, + 0.9997409582138062, + 0.999666154384613, + 0.9997338652610779, + 0.9998133182525635, + 0.9996403455734253, + 0.9996040463447571, + 0.9996645450592041, + 0.9995990991592407, + 0.9996405839920044, + 0.9995697140693665, + 0.9996727108955383, + 0.9995875358581543, + 0.9996808767318726, + 0.9995706081390381, + 0.9996305108070374, + 0.9995162487030029, + 0.9996678829193115, + 0.9995405673980713, + 0.9996769428253174, + 0.999527096748352, + 0.9997521638870239, + 0.9998356103897095, + 0.9998087882995605, + 0.9997869729995728, + 0.9997954368591309, + 0.9997106194496155, + 0.9997897148132324, + 0.9996377229690552, + 0.9996221661567688, + 0.9993540048599243, + 0.9995265007019043, + 0.999187707901001, + 0.9982263445854187, + 0.9977731704711914, + 0.998186469078064, + 0.9975561499595642, + 0.998430609703064, + 0.997812807559967, + 0.9999798536300659 + ], + "feedback_forward_norm_ratio": [ + 0.999843418598175, + 1.001550555229187, + 0.9994851350784302, + 1.0000321865081787, + 0.9982244968414307, + 1.001978874206543, + 0.9991020560264587, + 1.0019161701202393, + 0.998595118522644, + 1.0026040077209473, + 0.9997974634170532, + 1.001760721206665, + 0.9984849691390991, + 1.002119779586792, + 0.9994176626205444, + 1.0011266469955444, + 0.9993706941604614, + 1.0008323192596436, + 0.9987615346908569, + 1.000525712966919, + 0.9991268515586853, + 1.0014923810958862, + 0.998845636844635, + 1.001668095588684, + 0.9993511438369751, + 1.0019927024841309, + 0.9989683628082275, + 1.0017509460449219, + 0.9990468621253967, + 1.0016542673110962, + 0.9991765022277832, + 1.0017874240875244, + 0.9983173608779907, + 1.0019989013671875, + 0.9993149638175964, + 1.002145528793335, + 0.9993314146995544, + 1.0003092288970947, + 0.999877393245697, + 1.0006320476531982, + 0.99971604347229, + 1.000820279121399, + 0.999618649482727, + 1.0011359453201294, + 0.9993308186531067, + 1.0016674995422363, + 0.9991656541824341, + 1.002214789390564, + 0.9976616501808167, + 1.003898024559021, + 0.9978560209274292, + 1.0033684968948364, + 0.9981748461723328, + 1.0033618211746216, + 1.0017030239105225 + ], + "feedback_forward_relative_error": [ + 0.029293466359376907, + 0.03028717078268528, + 0.03443378582596779, + 0.03477523848414421, + 0.03089556284248829, + 0.03452856093645096, + 0.029456177726387978, + 0.03072730265557766, + 0.037744998931884766, + 0.03969145938754082, + 0.025676092132925987, + 0.027266312390565872, + 0.030022840946912766, + 0.030702665448188782, + 0.022397220134735107, + 0.026623448356986046, + 0.022760650143027306, + 0.025860438123345375, + 0.02309184893965721, + 0.019329769536852837, + 0.026822533458471298, + 0.028197942301630974, + 0.025912407785654068, + 0.02838928811252117, + 0.02681521326303482, + 0.029431317001581192, + 0.02559211105108261, + 0.02880263142287731, + 0.0252666138112545, + 0.0293738953769207, + 0.027186011895537376, + 0.03118078038096428, + 0.025803690776228905, + 0.030405156314373016, + 0.025418713688850403, + 0.03085964545607567, + 0.022268714383244514, + 0.01814079098403454, + 0.01955130510032177, + 0.020656032487750053, + 0.020225821062922478, + 0.024079715833067894, + 0.02050395868718624, + 0.026955166831612587, + 0.027489952743053436, + 0.036010999232530594, + 0.030774997547268867, + 0.0404089093208313, + 0.05953606218099594, + 0.06697878241539001, + 0.06019625440239906, + 0.07011085003614426, + 0.05600409582257271, + 0.06633620709180832, + 0.0065770866349339485 + ], + "innovation_negative_gradient_cosine": [ + 0.9994934797286987, + 0.9993982315063477, + 0.9994956254959106, + 0.9992268085479736, + 0.9994954466819763, + 0.9991641044616699, + 0.9994995594024658, + 0.9993889927864075, + 0.9995012283325195, + 0.9991922974586487, + 0.9995108246803284, + 0.9993957281112671, + 0.9995231628417969, + 0.9993530511856079, + 0.9995372295379639, + 0.9994377493858337, + 0.999556839466095, + 0.9994824528694153, + 0.9995675086975098, + 0.999570369720459, + 0.9995906352996826, + 0.99944007396698, + 0.9995976090431213, + 0.9994415044784546, + 0.9996044635772705, + 0.9994592070579529, + 0.9996154308319092, + 0.9994706511497498, + 0.9996240735054016, + 0.9995054602622986, + 0.9996492862701416, + 0.9995315074920654, + 0.9996607303619385, + 0.9996076822280884, + 0.999669075012207, + 0.9996030926704407, + 0.9996864795684814, + 0.9997321367263794, + 0.9997022747993469, + 0.9996989369392395, + 0.9997233748435974, + 0.9996992349624634, + 0.9997425675392151, + 0.9997498989105225, + 0.9997432231903076, + 0.9997694492340088, + 0.9997261166572571, + 0.9996894001960754, + 0.9996793866157532, + 0.999620258808136, + 0.9997217059135437, + 0.9996902346611023, + 0.9997488260269165, + 0.9997227787971497, + 0.735634446144104 + ], + "instruction_negative_gradient_cosine": [ + 0.9994934797286987, + 0.9993981719017029, + 0.9994956254959106, + 0.9992268085479736, + 0.9994955062866211, + 0.9991641044616699, + 0.9994996190071106, + 0.9993889331817627, + 0.9995012283325195, + 0.9991923570632935, + 0.9995107650756836, + 0.9993957281112671, + 0.9995231628417969, + 0.9993530511856079, + 0.9995372295379639, + 0.9994377493858337, + 0.999556839466095, + 0.9994824528694153, + 0.999567449092865, + 0.999570369720459, + 0.9995905160903931, + 0.99944007396698, + 0.9995976090431213, + 0.9994415044784546, + 0.9996045827865601, + 0.9994591474533081, + 0.9996154308319092, + 0.9994706511497498, + 0.9996240139007568, + 0.9995054602622986, + 0.9996492862701416, + 0.9995315074920654, + 0.9996607303619385, + 0.9996076226234436, + 0.999669075012207, + 0.9996031522750854, + 0.9996864795684814, + 0.9997321367263794, + 0.9997022747993469, + 0.9996988773345947, + 0.9997233152389526, + 0.9996992349624634, + 0.9997426271438599, + 0.9997498989105225, + 0.9997432231903076, + 0.999769389629364, + 0.9997261762619019, + 0.9996894001960754, + 0.999679446220398, + 0.999620258808136, + 0.9997217059135437, + 0.9996902346611023, + 0.9997488260269165, + 0.9997228384017944, + 0.742300808429718 + ], + "matched_negative_gradient_cosine": [ + 0.009145081043243408, + 0.020712576806545258, + 0.01115537527948618, + 0.029118096455931664, + 0.019411785528063774, + 0.05053672194480896, + 0.02483413927257061, + 0.017403408885002136, + 0.02556869015097618, + 0.05258864909410477, + 0.03975364565849304, + 0.05170352756977081, + 0.03912960737943649, + 0.039501167833805084, + 0.04481004923582077, + 0.0700264498591423, + 0.03045046329498291, + 0.005845815874636173, + 0.0425560437142849, + 0.038321420550346375, + 0.06834159791469574, + 0.13859328627586365, + 0.08878348022699356, + 0.12117676436901093, + 0.099797822535038, + 0.190749853849411, + 0.10534648597240448, + 0.19740569591522217, + 0.1054484024643898, + 0.2672116756439209, + 0.08820351958274841, + 0.2515257000923157, + 0.07720845937728882, + 0.2144172489643097, + 0.09138260036706924, + 0.28180715441703796, + 0.0978660061955452, + 0.08489158749580383, + 0.1037803590297699, + 0.27369028329849243, + 0.13156411051750183, + 0.3883955478668213, + 0.11124899238348007, + 0.3731602132320404, + 0.1306922882795334, + 0.34593671560287476, + 0.0874105766415596, + 0.2596692442893982, + 0.02722877264022827, + 0.07590722292661667, + 0.08146192133426666, + 0.17924267053604126, + 0.1025547906756401, + 0.14632907509803772, + 0.114814393222332 + ], + "max_norm_match_direction_error": 0.3530300259590149, + "max_norm_match_relative_error": 1.6793526924629987e-07, + "mean_feedback_forward_cosine": 0.999431696805087, + "mean_feedback_forward_relative_error": 0.03134233936328779, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.14260973036289215, + "max_absolute_post_projection_soma_slope": 4.247462115358758e-08, + "max_absolute_pre_projection_soma_slope": 0.14260973036289215, + "max_positive_post_projection_soma_slope": 1.9167110565376788e-08, + "min_post_projection_soma_slope": -4.247462115358758e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4350122690998085e-08, + "pre_projection_traffic_rms_ratio": 0.008812519354754533 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.00881251935472426, + "raw_negative_gradient_cosine": [ + 0.009145081974565983, + 0.020712576806545258, + 0.01115537527948618, + 0.029118096455931664, + 0.019411787390708923, + 0.05053672194480896, + 0.02483413740992546, + 0.017403408885002136, + 0.02556869015097618, + 0.05258864909410477, + 0.03975364565849304, + 0.05170352756977081, + 0.03912960737943649, + 0.039501167833805084, + 0.04481004923582077, + 0.0700264498591423, + 0.03045046329498291, + 0.0058458163402974606, + 0.0425560399889946, + 0.038321420550346375, + 0.06834159046411514, + 0.13859328627586365, + 0.08878347277641296, + 0.12117677181959152, + 0.09979782998561859, + 0.190749853849411, + 0.10534648597240448, + 0.19740569591522217, + 0.1054484024643898, + 0.2672116756439209, + 0.08820351958274841, + 0.2515257000923157, + 0.07720845937728882, + 0.2144172489643097, + 0.09138260036706924, + 0.28180715441703796, + 0.0978659987449646, + 0.08489158749580383, + 0.1037803590297699, + 0.2736903131008148, + 0.13156409561634064, + 0.3883955478668213, + 0.11124899983406067, + 0.3731602132320404, + 0.1306922882795334, + 0.34593671560287476, + 0.0874105766415596, + 0.2596692442893982, + 0.02722877264022827, + 0.07590721547603607, + 0.08146192133426666, + 0.17924265563488007, + 0.1025547981262207, + 0.14632906019687653, + 0.11466988921165466 + ], + "teaching_negative_gradient_cosine": [ + 0.9994934797286987, + 0.9993982315063477, + 0.9994956254959106, + 0.9992268085479736, + 0.9994954466819763, + 0.9991641044616699, + 0.9994995594024658, + 0.9993889927864075, + 0.9995012283325195, + 0.9991922974586487, + 0.9995108246803284, + 0.9993957281112671, + 0.9995231628417969, + 0.9993530511856079, + 0.9995372295379639, + 0.9994377493858337, + 0.999556839466095, + 0.9994824528694153, + 0.9995675086975098, + 0.999570369720459, + 0.9995906352996826, + 0.99944007396698, + 0.9995976090431213, + 0.9994415044784546, + 0.9996044635772705, + 0.9994592070579529, + 0.9996154308319092, + 0.9994706511497498, + 0.9996240735054016, + 0.9995054602622986, + 0.9996492862701416, + 0.9995315074920654, + 0.9996607303619385, + 0.9996076822280884, + 0.999669075012207, + 0.9996030926704407, + 0.9996864795684814, + 0.9997321367263794, + 0.9997022747993469, + 0.9996989369392395, + 0.9997233748435974, + 0.9996992349624634, + 0.9997425675392151, + 0.9997498989105225, + 0.9997432231903076, + 0.9997694492340088, + 0.9997261166572571, + 0.9996894001960754, + 0.9996793866157532, + 0.999620258808136, + 0.9997217059135437, + 0.9996902346611023, + 0.9997488260269165, + 0.9997227787971497, + 0.735634446144104 + ], + "traffic_instruction_rms_ratio": 21.865208705270707, + "used_negative_gradient_cosine": [ + 0.9994934797286987, + 0.9993982315063477, + 0.9994956254959106, + 0.9992268085479736, + 0.9994954466819763, + 0.9991641044616699, + 0.9994995594024658, + 0.9993889927864075, + 0.9995012283325195, + 0.9991922974586487, + 0.9995108246803284, + 0.9993957281112671, + 0.9995231628417969, + 0.9993530511856079, + 0.9995372295379639, + 0.9994377493858337, + 0.999556839466095, + 0.9994824528694153, + 0.9995675086975098, + 0.999570369720459, + 0.9995906352996826, + 0.99944007396698, + 0.9995976090431213, + 0.9994415044784546, + 0.9996044635772705, + 0.9994592070579529, + 0.9996154308319092, + 0.9994706511497498, + 0.9996240735054016, + 0.9995054602622986, + 0.9996492862701416, + 0.9995315074920654, + 0.9996607303619385, + 0.9996076822280884, + 0.999669075012207, + 0.9996030926704407, + 0.9996864795684814, + 0.9997321367263794, + 0.9997022747993469, + 0.9996989369392395, + 0.9997233748435974, + 0.9996992349624634, + 0.9997425675392151, + 0.9997498989105225, + 0.9997432231903076, + 0.9997694492340088, + 0.9997261166572571, + 0.9996894001960754, + 0.9996793866157532, + 0.999620258808136, + 0.9997217059135437, + 0.9996902346611023, + 0.9997488260269165, + 0.9997227787971497, + 0.735634446144104 + ], + "wall_s": 0.34674644470214844 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3731249570846558, + "mean_feedback_forward_cosine": 0.24383907169103622, + "mean_feedback_forward_relative_error": 1.2229380065744573, + "min_feedback_forward_cosine": 0.06914597004652023 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0011735415065878042, + "instruction_rms": 0.0011735415065781127, + "raw_apical_rms": 0.03970981731480759, + "teaching_rms": 0.0011735415065878042, + "traffic_rms": 0.039589583590077214 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260977506637573, + "maximum_absolute_post_projection_soma_slope": 8.299865328353917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.0173287463056015e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014547473336261161, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.4184074737548826 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3461002111434937, + "mean_feedback_forward_cosine": 0.2696815949949351, + "mean_feedback_forward_relative_error": 1.2026612693613226, + "min_feedback_forward_cosine": 0.1001327857375145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0011514765673950236, + "instruction_rms": 0.001151476567255391, + "raw_apical_rms": 0.045794743736385236, + "teaching_rms": 0.0011514765673950236, + "traffic_rms": 0.0457799909076458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260976016521454, + "maximum_absolute_post_projection_soma_slope": 5.1173611126387186e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8548287900272433e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017615210458279712, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.762520817604065 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3408757448196411, + "mean_feedback_forward_cosine": 0.3161876526745883, + "mean_feedback_forward_relative_error": 1.1650627585974607, + "min_feedback_forward_cosine": 0.10489001870155334 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0016549438746414546, + "instruction_rms": 0.0016549438748159956, + "raw_apical_rms": 0.05151741165488746, + "teaching_rms": 0.0016549438746414546, + "traffic_rms": 0.051490281252015324 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.5940943493860686e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.815517811552671e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016994714176783932, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.512260643157959 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3316549062728882, + "mean_feedback_forward_cosine": 0.3749822615222497, + "mean_feedback_forward_relative_error": 1.1154337584972382, + "min_feedback_forward_cosine": 0.11218077689409256 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002275252031508896, + "instruction_rms": 0.0022752520312655663, + "raw_apical_rms": 0.0590345048903686, + "teaching_rms": 0.002275252031508896, + "traffic_rms": 0.058990198245991 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.231281363649032e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7644759047707248e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017628377766492665, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.2438516765213012 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3258413076400757, + "mean_feedback_forward_cosine": 0.42951178672638807, + "mean_feedback_forward_relative_error": 1.0656936445019463, + "min_feedback_forward_cosine": 0.1209106519818306 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002643699691281606, + "instruction_rms": 0.002643699691855183, + "raw_apical_rms": 0.06510588726889914, + "teaching_rms": 0.002643699691281606, + "traffic_rms": 0.06505173114730729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.0512151855409684e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.722713432202185e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01723250722562036, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.0136612246322632 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3194870948791504, + "mean_feedback_forward_cosine": 0.4723150819540024, + "mean_feedback_forward_relative_error": 1.0242296571081335, + "min_feedback_forward_cosine": 0.13022229075431824 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027426750034518807, + "instruction_rms": 0.002742675002665937, + "raw_apical_rms": 0.06990917648948562, + "teaching_rms": 0.0027426750034518807, + "traffic_rms": 0.06985486620014025 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.57405597842353e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6861699679115742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01685556838342217, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.8622315446090698 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3131864070892334, + "mean_feedback_forward_cosine": 0.5075092513452877, + "mean_feedback_forward_relative_error": 0.9886099425229159, + "min_feedback_forward_cosine": 0.13890588283538818 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027565688461020418, + "instruction_rms": 0.0027565688461008665, + "raw_apical_rms": 0.0732243208573209, + "teaching_rms": 0.0027565688461020418, + "traffic_rms": 0.0731720057025109 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.1846366372055854e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6607284779938954e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016290882085779498, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.7612383555221558 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3064661026000977, + "mean_feedback_forward_cosine": 0.5395994912494313, + "mean_feedback_forward_relative_error": 0.9543644444508986, + "min_feedback_forward_cosine": 0.14829930663108826 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029274980586611993, + "instruction_rms": 0.00292749805821342, + "raw_apical_rms": 0.07602529356063284, + "teaching_rms": 0.0029274980586611993, + "traffic_rms": 0.07596838769331994 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.0723592437407206e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6519811884155084e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015116814756271498, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.6853854769706726 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2993086576461792, + "mean_feedback_forward_cosine": 0.5678306186741049, + "mean_feedback_forward_relative_error": 0.9229284010150216, + "min_feedback_forward_cosine": 0.15878009796142578 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002925969400428821, + "instruction_rms": 0.0029259694004951952, + "raw_apical_rms": 0.07765925269197753, + "teaching_rms": 0.002925969400428821, + "traffic_rms": 0.07760362227067051 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.743388780705573e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6339202143355274e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014225178135468407, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.6295745374870301 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2910151481628418, + "mean_feedback_forward_cosine": 0.593825030055913, + "mean_feedback_forward_relative_error": 0.8929989077828148, + "min_feedback_forward_cosine": 0.170100599527359 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002987865834335705, + "instruction_rms": 0.002987865833552953, + "raw_apical_rms": 0.07899781976903276, + "teaching_rms": 0.002987865834335705, + "traffic_rms": 0.07894057704038218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.923046503790829e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6315237383022224e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015077701846357223, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.5860557368278504 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2823660373687744, + "mean_feedback_forward_cosine": 0.6178091233426874, + "mean_feedback_forward_relative_error": 0.8643731014295057, + "min_feedback_forward_cosine": 0.18129217624664307 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030597157851392515, + "instruction_rms": 0.0030597157859018915, + "raw_apical_rms": 0.0804539761410236, + "teaching_rms": 0.0030597157851392515, + "traffic_rms": 0.08039518505816531 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.720018236843316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6275350507650064e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014758505621887503, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5491757737350463 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2722746133804321, + "mean_feedback_forward_cosine": 0.6401401311159134, + "mean_feedback_forward_relative_error": 0.8367454994808544, + "min_feedback_forward_cosine": 0.19456332921981812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030957983647721295, + "instruction_rms": 0.0030957983640941145, + "raw_apical_rms": 0.08161926230133791, + "teaching_rms": 0.0030957983647721295, + "traffic_rms": 0.08155977108612134 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.139646930274466e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.609164351010329e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015799664481893422, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5175455815505982 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2608096599578857, + "mean_feedback_forward_cosine": 0.660786991498687, + "mean_feedback_forward_relative_error": 0.81037946424701, + "min_feedback_forward_cosine": 0.20936188101768494 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030781235354573567, + "instruction_rms": 0.003078123537057369, + "raw_apical_rms": 0.08337431725574489, + "teaching_rms": 0.0030781235354573567, + "traffic_rms": 0.08331677915938523 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.17742826389167e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.604457836068378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015449845241321073, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.48272350957870486 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.248460292816162, + "mean_feedback_forward_cosine": 0.6798852021043951, + "mean_feedback_forward_relative_error": 0.7852349094369194, + "min_feedback_forward_cosine": 0.22457371652126312 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003100339211295203, + "instruction_rms": 0.0031003392114249467, + "raw_apical_rms": 0.08494837575713804, + "teaching_rms": 0.003100339211295203, + "traffic_rms": 0.08489106844723374 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.217096088472317e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5957172868956847e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01433126936808082, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.4695433413505554 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2356730699539185, + "mean_feedback_forward_cosine": 0.6979092966426502, + "mean_feedback_forward_relative_error": 0.7607278384945609, + "min_feedback_forward_cosine": 0.24089950323104858 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030866216165242872, + "instruction_rms": 0.0030866216169087583, + "raw_apical_rms": 0.08555312030032036, + "teaching_rms": 0.0030866216165242872, + "traffic_rms": 0.08549662477664306 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 3.915413770982923e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6069746057882416e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014215939041889511, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.44431737590789794 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.221461534500122, + "mean_feedback_forward_cosine": 0.7151031217791818, + "mean_feedback_forward_relative_error": 0.7367805808782577, + "min_feedback_forward_cosine": 0.2581232786178589 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003150088060700533, + "instruction_rms": 0.0031500880620992425, + "raw_apical_rms": 0.08668502895686502, + "teaching_rms": 0.003150088060700533, + "traffic_rms": 0.08662691618474584 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.427119648653388e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5926292691735105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01514278446834455, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4272077713012695 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2064054012298584, + "mean_feedback_forward_cosine": 0.7313121627677571, + "mean_feedback_forward_relative_error": 0.7133815402334387, + "min_feedback_forward_cosine": 0.27655303478240967 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031676989314134937, + "instruction_rms": 0.0031676989313967753, + "raw_apical_rms": 0.08779876860020117, + "teaching_rms": 0.0031676989314134937, + "traffic_rms": 0.08774073357977569 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.891075988666671e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5845491407552892e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015564226046954089, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.41010695198059083 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1892849206924438, + "mean_feedback_forward_cosine": 0.7461305732076818, + "mean_feedback_forward_relative_error": 0.6915872013027018, + "min_feedback_forward_cosine": 0.29781749844551086 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031125275056129464, + "instruction_rms": 0.0031125275041885103, + "raw_apical_rms": 0.08895119841734302, + "teaching_rms": 0.0031125275056129464, + "traffic_rms": 0.0888956884749632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.4969496570956835e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.576501189309585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015773583572524803, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.3859537725162506 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1772735118865967, + "mean_feedback_forward_cosine": 0.760148654200814, + "mean_feedback_forward_relative_error": 0.6704453896392476, + "min_feedback_forward_cosine": 0.3165215253829956 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030907806372996423, + "instruction_rms": 0.003090780635865599, + "raw_apical_rms": 0.08980578390375954, + "teaching_rms": 0.0030907806372996423, + "traffic_rms": 0.08975168761137799 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.059389789925262e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5769334612786096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015363409185099226, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.37772658000946047 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1634541749954224, + "mean_feedback_forward_cosine": 0.7733385806733911, + "mean_feedback_forward_relative_error": 0.649839403683489, + "min_feedback_forward_cosine": 0.33794498443603516 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030918706624569413, + "instruction_rms": 0.003091870662278607, + "raw_apical_rms": 0.09041249729278336, + "teaching_rms": 0.0030918706624569413, + "traffic_rms": 0.09035864335020174 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.208373714069239e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5827124977612797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0155473242950958, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.3625246440696716 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1474006175994873, + "mean_feedback_forward_cosine": 0.7861961467699571, + "mean_feedback_forward_relative_error": 0.6294605274092068, + "min_feedback_forward_cosine": 0.35897910594940186 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003063784300643496, + "instruction_rms": 0.0030637843012309806, + "raw_apical_rms": 0.09130127876237354, + "teaching_rms": 0.003063784300643496, + "traffic_rms": 0.0912489663455528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.200717012598943e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5743717851999996e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015155502015739801, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.3535980351257324 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1316689252853394, + "mean_feedback_forward_cosine": 0.7981834920969877, + "mean_feedback_forward_relative_error": 0.6098968860777941, + "min_feedback_forward_cosine": 0.3776004910469055 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030198313112314826, + "instruction_rms": 0.003019831310273336, + "raw_apical_rms": 0.09212686628102663, + "teaching_rms": 0.0030198313112314826, + "traffic_rms": 0.09207642995162349 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.472953645835332e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5720831404804856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01448153092935823, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.33958834842681884 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1145658493041992, + "mean_feedback_forward_cosine": 0.8099844981323588, + "mean_feedback_forward_relative_error": 0.5901308189738881, + "min_feedback_forward_cosine": 0.39721959829330444 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031095366915210847, + "instruction_rms": 0.003109536691648369, + "raw_apical_rms": 0.09278682963769078, + "teaching_rms": 0.0031095366915210847, + "traffic_rms": 0.0927337075329886 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.945604269503747e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5798435995436544e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013738821898449874, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3287642791175842 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0978220701217651, + "mean_feedback_forward_cosine": 0.8209721792827953, + "mean_feedback_forward_relative_error": 0.5711639325727116, + "min_feedback_forward_cosine": 0.415591835975647 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003069652432316055, + "instruction_rms": 0.0030696524335324443, + "raw_apical_rms": 0.09348609713081485, + "teaching_rms": 0.003069652432316055, + "traffic_rms": 0.09343481385526388 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.79805731099259e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583048674027514e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013846190205704642, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.32195955295562745 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0809894800186157, + "mean_feedback_forward_cosine": 0.8312957118857991, + "mean_feedback_forward_relative_error": 0.5528859303756194, + "min_feedback_forward_cosine": 0.43454086780548096 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030282144491123974, + "instruction_rms": 0.0030282144488740907, + "raw_apical_rms": 0.09397255677229363, + "teaching_rms": 0.0030282144491123974, + "traffic_rms": 0.09392287676552098 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.567330501638935e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5723765798822936e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014164544501722831, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.31536266344070435 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0646122694015503, + "mean_feedback_forward_cosine": 0.8409607627175071, + "mean_feedback_forward_relative_error": 0.5353314527056434, + "min_feedback_forward_cosine": 0.45218151807785034 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030546850440231997, + "instruction_rms": 0.0030546850445190483, + "raw_apical_rms": 0.09485936188926189, + "teaching_rms": 0.0030546850440231997, + "traffic_rms": 0.09480915069210738 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.25003985735384e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5737309332038156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014189474440568207, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.30200243107795716 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0441334247589111, + "mean_feedback_forward_cosine": 0.8504141520370136, + "mean_feedback_forward_relative_error": 0.5177891796285455, + "min_feedback_forward_cosine": 0.47327572107315063 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029530534007487356, + "instruction_rms": 0.0029530533994777913, + "raw_apical_rms": 0.09547829037433414, + "teaching_rms": 0.0029530534007487356, + "traffic_rms": 0.09543182297431305 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.601877799994327e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5757670171666155e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014016359018883024, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.29573998839378357 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.023980975151062, + "mean_feedback_forward_cosine": 0.8593124503439123, + "mean_feedback_forward_relative_error": 0.5008702359416268, + "min_feedback_forward_cosine": 0.4940739572048187 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029514427115604, + "instruction_rms": 0.002951442711149523, + "raw_apical_rms": 0.09582475724050606, + "teaching_rms": 0.0029514427115604, + "traffic_rms": 0.09577826114355201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.924273776576229e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570709884971202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013079197034457856, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.2887161243343353 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0031672716140747, + "mean_feedback_forward_cosine": 0.8676217436790467, + "mean_feedback_forward_relative_error": 0.48474472056735646, + "min_feedback_forward_cosine": 0.5147253274917603 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029150398356783407, + "instruction_rms": 0.0029150398362023677, + "raw_apical_rms": 0.09664069460423406, + "teaching_rms": 0.0029150398356783407, + "traffic_rms": 0.09659575359011043 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.6425445649210815e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5729916984751467e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012195946340785711, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.27999820298194883 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9808943867683411, + "mean_feedback_forward_cosine": 0.8758031595836986, + "mean_feedback_forward_relative_error": 0.46841765655712647, + "min_feedback_forward_cosine": 0.536214292049408 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029537241736217724, + "instruction_rms": 0.0029537241733964188, + "raw_apical_rms": 0.09727864540573213, + "teaching_rms": 0.0029537241736217724, + "traffic_rms": 0.09723266629350111 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.832315335714156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5707884872185697e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012995043499540572, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.27962089896678927 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9596076607704163, + "mean_feedback_forward_cosine": 0.8834901343692433, + "mean_feedback_forward_relative_error": 0.45259491096843374, + "min_feedback_forward_cosine": 0.556028425693512 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029701928797881633, + "instruction_rms": 0.0029701928798994575, + "raw_apical_rms": 0.09747904508748882, + "teaching_rms": 0.0029701928797881633, + "traffic_rms": 0.09743268181722847 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.079094561073362e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.573537308775561e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014137449213955057, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.26991109534263613 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9381541013717651, + "mean_feedback_forward_cosine": 0.8907324595884843, + "mean_feedback_forward_relative_error": 0.43744216046550055, + "min_feedback_forward_cosine": 0.5759724974632263 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028884212164978643, + "instruction_rms": 0.0028884212168105643, + "raw_apical_rms": 0.09820832805707319, + "teaching_rms": 0.0028884212164978643, + "traffic_rms": 0.0981648624058883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.911571321348674e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567879417193024e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01381901980511141, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.26592113607406614 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9172576665878296, + "mean_feedback_forward_cosine": 0.8973997538739985, + "mean_feedback_forward_relative_error": 0.4229122679341923, + "min_feedback_forward_cosine": 0.5946725606918335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029044720991591643, + "instruction_rms": 0.002904472099249637, + "raw_apical_rms": 0.09838487407131767, + "teaching_rms": 0.0029044720991591643, + "traffic_rms": 0.09834096655182067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.061917190436361e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.575228782406986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013126473028488874, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.260477105550766 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8941831588745117, + "mean_feedback_forward_cosine": 0.9040087602355263, + "mean_feedback_forward_relative_error": 0.4081582895734093, + "min_feedback_forward_cosine": 0.6147831082344055 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002953903341352604, + "instruction_rms": 0.002953903341786973, + "raw_apical_rms": 0.09907064424544468, + "teaching_rms": 0.002953903341352604, + "traffic_rms": 0.0990255710373813 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.6642157630903966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.573925055685662e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013614832887721176, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.2544909772396088 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8725862503051758, + "mean_feedback_forward_cosine": 0.9099419561299411, + "mean_feedback_forward_relative_error": 0.39455513561313804, + "min_feedback_forward_cosine": 0.6333650350570679 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002872591562396657, + "instruction_rms": 0.0028725915623755796, + "raw_apical_rms": 0.09995197401773434, + "teaching_rms": 0.002872591562396657, + "traffic_rms": 0.09990960024613042 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.920808542394298e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568166077926738e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01302297976245306, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.2500183685874939 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8505612015724182, + "mean_feedback_forward_cosine": 0.9156503211368214, + "mean_feedback_forward_relative_error": 0.381031512807716, + "min_feedback_forward_cosine": 0.6518673896789551 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002906328946952155, + "instruction_rms": 0.002906328947095265, + "raw_apical_rms": 0.10041930511289587, + "teaching_rms": 0.002906328946952155, + "traffic_rms": 0.10037613554743059 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.028780484892195e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5715670189477694e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013175141137930668, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.2429008005809784 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8272808194160461, + "mean_feedback_forward_cosine": 0.9212871724909002, + "mean_feedback_forward_relative_error": 0.36750791370868685, + "min_feedback_forward_cosine": 0.6705191135406494 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00287719147551969, + "instruction_rms": 0.0028771914749242114, + "raw_apical_rms": 0.10080540078566559, + "teaching_rms": 0.00287719147551969, + "traffic_rms": 0.10076341136521186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.819430353743883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.569205869181346e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013809687399581291, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24378767097473145 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8031185865402222, + "mean_feedback_forward_cosine": 0.9264169758016413, + "mean_feedback_forward_relative_error": 0.35485375090078874, + "min_feedback_forward_cosine": 0.6895716786384583 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028588993115555073, + "instruction_rms": 0.002858899311487647, + "raw_apical_rms": 0.10108642354232243, + "teaching_rms": 0.0028588993115555073, + "traffic_rms": 0.10104492207791024 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.419724485340339e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5810795484351968e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014169441723367232, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.23416394399642945 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7798122763633728, + "mean_feedback_forward_cosine": 0.9313094930215315, + "mean_feedback_forward_relative_error": 0.3424241114746441, + "min_feedback_forward_cosine": 0.7070998549461365 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029049965101521583, + "instruction_rms": 0.0029049965094200786, + "raw_apical_rms": 0.10150485219888974, + "teaching_rms": 0.0029049965101521583, + "traffic_rms": 0.10146215450053 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.224089332818039e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5778804391883247e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013184222199463207, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23258170228004454 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7568359971046448, + "mean_feedback_forward_cosine": 0.9358303720300848, + "mean_feedback_forward_relative_error": 0.3304865131323988, + "min_feedback_forward_cosine": 0.7243583798408508 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002866647865348375, + "instruction_rms": 0.002866647865046059, + "raw_apical_rms": 0.10214047763590728, + "teaching_rms": 0.002866647865348375, + "traffic_rms": 0.10209911196706857 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.137473380045776e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5776476833254052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013452820860951592, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.22841290390968322 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7354384660720825, + "mean_feedback_forward_cosine": 0.9401394443078475, + "mean_feedback_forward_relative_error": 0.318697267499837, + "min_feedback_forward_cosine": 0.7395153641700745 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028548351238196477, + "instruction_rms": 0.0028548351233666593, + "raw_apical_rms": 0.10259300864405378, + "teaching_rms": 0.0028548351238196477, + "traffic_rms": 0.10255232180151314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.349424973630448e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.567145283431142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012975653140872852, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.22398219423294066 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7125124335289001, + "mean_feedback_forward_cosine": 0.944255594773726, + "mean_feedback_forward_relative_error": 0.3072501927614212, + "min_feedback_forward_cosine": 0.7553398609161377 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00281822510065321, + "instruction_rms": 0.0028182251002738684, + "raw_apical_rms": 0.10282596074602184, + "teaching_rms": 0.00281822510065321, + "traffic_rms": 0.10278618141043108 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.371536270284196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5667496093405423e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012975807213933805, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.22099732969284058 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6898049116134644, + "mean_feedback_forward_cosine": 0.9482305364175276, + "mean_feedback_forward_relative_error": 0.2957717233083465, + "min_feedback_forward_cosine": 0.7705265283584595 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028973856140803193, + "instruction_rms": 0.002897385612762822, + "raw_apical_rms": 0.10320477788478243, + "teaching_rms": 0.0028973856140803193, + "traffic_rms": 0.1031627705074429 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.912096412430401e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5739904923923746e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012765481672362317, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22500161635398866 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6682578921318054, + "mean_feedback_forward_cosine": 0.9518158500844782, + "mean_feedback_forward_relative_error": 0.28499872630292716, + "min_feedback_forward_cosine": 0.7844188213348389 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028497793048575374, + "instruction_rms": 0.0028497793040572184, + "raw_apical_rms": 0.1034023438337447, + "teaching_rms": 0.0028497793048575374, + "traffic_rms": 0.10336207453437073 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.125180528395504e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5755280397364038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011794718002530893, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.21802832321166993 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.645577073097229, + "mean_feedback_forward_cosine": 0.9553018710829995, + "mean_feedback_forward_relative_error": 0.27436178543350914, + "min_feedback_forward_cosine": 0.7986353635787964 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002837690433715268, + "instruction_rms": 0.00283769043323183, + "raw_apical_rms": 0.1039405543094929, + "teaching_rms": 0.002837690433715268, + "traffic_rms": 0.10390072790730881 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 1.0120236026978091e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.588044016207537e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013027161154706089, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.21780684286355972 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6229442358016968, + "mean_feedback_forward_cosine": 0.9584865526719527, + "mean_feedback_forward_relative_error": 0.264249609681693, + "min_feedback_forward_cosine": 0.812308669090271 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027915486086066816, + "instruction_rms": 0.0027915486094637464, + "raw_apical_rms": 0.1043951508394249, + "teaching_rms": 0.0027915486086066816, + "traffic_rms": 0.10435679141347277 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.430759924569429e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5843794422713387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01413915311893475, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.21099490920066832 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6019085049629211, + "mean_feedback_forward_cosine": 0.9614515900611877, + "mean_feedback_forward_relative_error": 0.2544229306280613, + "min_feedback_forward_cosine": 0.8246707320213318 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002835541800307012, + "instruction_rms": 0.0028355418005442667, + "raw_apical_rms": 0.10452683350498693, + "teaching_rms": 0.002835541800307012, + "traffic_rms": 0.10448723898247161 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.07220371762196e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5782247721053196e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013737600776080796, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.2115399778652191 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5820757746696472, + "mean_feedback_forward_cosine": 0.9641605323011225, + "mean_feedback_forward_relative_error": 0.2451463830741969, + "min_feedback_forward_cosine": 0.8359454274177551 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027966726106549624, + "instruction_rms": 0.0027966726105186097, + "raw_apical_rms": 0.10490332180376637, + "teaching_rms": 0.0027966726106549624, + "traffic_rms": 0.10486499001268913 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.094941961189761e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5791095635521253e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013813689956498888, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2097087073802948 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5615283846855164, + "mean_feedback_forward_cosine": 0.9667730266397649, + "mean_feedback_forward_relative_error": 0.23596984527327797, + "min_feedback_forward_cosine": 0.847258985042572 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002830530347047628, + "instruction_rms": 0.0028305303478731703, + "raw_apical_rms": 0.10527902830915171, + "teaching_rms": 0.002830530347047628, + "traffic_rms": 0.1052397913362817 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.577518346124634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5747393843846808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011416830738954288, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20351419405937196 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5412197709083557, + "mean_feedback_forward_cosine": 0.9692050034349615, + "mean_feedback_forward_relative_error": 0.22713793001391672, + "min_feedback_forward_cosine": 0.8579676151275635 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002778881056267558, + "instruction_rms": 0.002778881056337272, + "raw_apical_rms": 0.10612983317923944, + "teaching_rms": 0.002778881056267558, + "traffic_rms": 0.1060924508517643 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.599479718261136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.578457315249057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011651813818267524, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20360362627029419 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5217387080192566, + "mean_feedback_forward_cosine": 0.971462222662839, + "mean_feedback_forward_relative_error": 0.21857626539739697, + "min_feedback_forward_cosine": 0.8679684400558472 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028066155389162508, + "instruction_rms": 0.0028066155389387965, + "raw_apical_rms": 0.10641312279799853, + "teaching_rms": 0.0028066155389162508, + "traffic_rms": 0.1063749831403136 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.587332279650582e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585370674539895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0124907056244175, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20178700191497803 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5027214884757996, + "mean_feedback_forward_cosine": 0.9735671281814575, + "mean_feedback_forward_relative_error": 0.21029050282456657, + "min_feedback_forward_cosine": 0.8772383332252502 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002809373391872543, + "instruction_rms": 0.0028093733908111165, + "raw_apical_rms": 0.10666886586200684, + "teaching_rms": 0.002809373391872543, + "traffic_rms": 0.10663076199164984 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.233867194145205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5882390109250447e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014260951142897832, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.19983881056308747 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4841027557849884, + "mean_feedback_forward_cosine": 0.975510357726704, + "mean_feedback_forward_relative_error": 0.2023856523362073, + "min_feedback_forward_cosine": 0.886039137840271 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028057962914911465, + "instruction_rms": 0.0028057962918048855, + "raw_apical_rms": 0.10683032787935892, + "teaching_rms": 0.0028057962914911465, + "traffic_rms": 0.10679238232862305 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.958634995475222e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5890317067371668e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014156637493983154, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19527202021121978 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4683317542076111, + "mean_feedback_forward_cosine": 0.9772859768434005, + "mean_feedback_forward_relative_error": 0.19480349360541863, + "min_feedback_forward_cosine": 0.893265426158905 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028115105795083442, + "instruction_rms": 0.0028115105801477885, + "raw_apical_rms": 0.10708152791222089, + "teaching_rms": 0.0028115105795083442, + "traffic_rms": 0.10704334186256174 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.91136605723841e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583921775804431e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013944639376534002, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19435997552871703 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45155733823776245, + "mean_feedback_forward_cosine": 0.9789445183493874, + "mean_feedback_forward_relative_error": 0.18746368620883336, + "min_feedback_forward_cosine": 0.9007347822189331 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028361027151679115, + "instruction_rms": 0.002836102715629231, + "raw_apical_rms": 0.10769748967951297, + "teaching_rms": 0.0028361027151679115, + "traffic_rms": 0.10765901571608535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.738077746286763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.588367410410188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012556639159839879, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19328636796951293 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4338410198688507, + "mean_feedback_forward_cosine": 0.9804906996813687, + "mean_feedback_forward_relative_error": 0.18052111166444693, + "min_feedback_forward_cosine": 0.9082486033439636 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002753131319100576, + "instruction_rms": 0.0027531313183625497, + "raw_apical_rms": 0.10800181224971232, + "teaching_rms": 0.002753131319100576, + "traffic_rms": 0.10796563604551893 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.138017027979004e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5755848181236358e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012139309117807418, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.18370914353370665 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41759082674980164, + "mean_feedback_forward_cosine": 0.9819487799297679, + "mean_feedback_forward_relative_error": 0.17362151538783854, + "min_feedback_forward_cosine": 0.9149218797683716 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028445505072922407, + "instruction_rms": 0.0028445505074379336, + "raw_apical_rms": 0.1084722792705982, + "teaching_rms": 0.0028445505072922407, + "traffic_rms": 0.10843395931542137 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.160276365742902e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5753305957672444e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012058021925897739, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.1886521065092087 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4019881784915924, + "mean_feedback_forward_cosine": 0.9832683129744096, + "mean_feedback_forward_relative_error": 0.1671425577591766, + "min_feedback_forward_cosine": 0.9210941791534424 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027730928359734875, + "instruction_rms": 0.002773092835889914, + "raw_apical_rms": 0.10894468491127962, + "teaching_rms": 0.0027730928359734875, + "traffic_rms": 0.10890825987558718 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.118245548554114e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5913891035340608e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011319808873441906, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.18609938282966615 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38494738936424255, + "mean_feedback_forward_cosine": 0.9845792997967113, + "mean_feedback_forward_relative_error": 0.1605221392078833, + "min_feedback_forward_cosine": 0.9275520443916321 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028340585939216046, + "instruction_rms": 0.0028340585931943387, + "raw_apical_rms": 0.10898095063537308, + "teaching_rms": 0.0028340585939216046, + "traffic_rms": 0.10894289515238936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.121654022057555e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5863591984244917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010946963560757698, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18812277416229248 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.369772344827652, + "mean_feedback_forward_cosine": 0.9857169183817777, + "mean_feedback_forward_relative_error": 0.15450807017358867, + "min_feedback_forward_cosine": 0.9331113696098328 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027492148448285784, + "instruction_rms": 0.002749214845614018, + "raw_apical_rms": 0.10919187138781966, + "teaching_rms": 0.0027492148448285784, + "traffic_rms": 0.10915615532227238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.21795868735353e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5858688654600588e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011998612412698987, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.1817421946144104 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3549457788467407, + "mean_feedback_forward_cosine": 0.9868247433142229, + "mean_feedback_forward_relative_error": 0.14843868999318643, + "min_feedback_forward_cosine": 0.9383153915405273 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002795432337826108, + "instruction_rms": 0.0027954323376297168, + "raw_apical_rms": 0.10953554425308172, + "teaching_rms": 0.002795432337826108, + "traffic_rms": 0.10949868365423558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.403677753676675e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5811916112277436e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012504434126110155, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.18432221712112426 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3409314751625061, + "mean_feedback_forward_cosine": 0.9878174304962158, + "mean_feedback_forward_relative_error": 0.1427491165020249, + "min_feedback_forward_cosine": 0.9430403709411621 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027719022988168442, + "instruction_rms": 0.0027719022985481564, + "raw_apical_rms": 0.10980866853435771, + "teaching_rms": 0.0027719022988168442, + "traffic_rms": 0.10977260463852206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.02628071280742e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.59795019963987e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014288228928888626, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18277811943531036 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3275431990623474, + "mean_feedback_forward_cosine": 0.9887475967407227, + "mean_feedback_forward_relative_error": 0.13718989274718546, + "min_feedback_forward_cosine": 0.9473963975906372 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002786289477603806, + "instruction_rms": 0.002786289477717537, + "raw_apical_rms": 0.11035204121235341, + "teaching_rms": 0.002786289477603806, + "traffic_rms": 0.11031567598082853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.8033482019936855e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5830514883847327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012268488733917598, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18441451013565063 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3149729371070862, + "mean_feedback_forward_cosine": 0.9895734602754767, + "mean_feedback_forward_relative_error": 0.13212151005864142, + "min_feedback_forward_cosine": 0.9513528347015381 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002684091019023277, + "instruction_rms": 0.0026840910196919295, + "raw_apical_rms": 0.11074311543036824, + "teaching_rms": 0.002684091019023277, + "traffic_rms": 0.11070934474189535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.500812332888017e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5747356365169516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012162142801696387, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.17204994041204452 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3023730218410492, + "mean_feedback_forward_cosine": 0.9903738455338912, + "mean_feedback_forward_relative_error": 0.1269822800701315, + "min_feedback_forward_cosine": 0.9551512002944946 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028041684662593685, + "instruction_rms": 0.002804168465669102, + "raw_apical_rms": 0.11100716687586223, + "teaching_rms": 0.0028041684662593685, + "traffic_rms": 0.11097038758601827 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.003244917314078e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.590774364346127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01235985994375744, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1738403448152542 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2906089425086975, + "mean_feedback_forward_cosine": 0.9910957325588573, + "mean_feedback_forward_relative_error": 0.12216384509070354, + "min_feedback_forward_cosine": 0.9585556387901306 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744578840305552, + "instruction_rms": 0.002744578840193564, + "raw_apical_rms": 0.11135327303571363, + "teaching_rms": 0.002744578840305552, + "traffic_rms": 0.1113182206351424 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.252631834513522e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6017954393625452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012932227418574152, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.174180910050869 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2792268395423889, + "mean_feedback_forward_cosine": 0.9917706814679232, + "mean_feedback_forward_relative_error": 0.11749785826964812, + "min_feedback_forward_cosine": 0.9617061018943787 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027002174870360404, + "instruction_rms": 0.0027002174872921107, + "raw_apical_rms": 0.11190783294569717, + "teaching_rms": 0.0027002174870360404, + "traffic_rms": 0.11187410778516534 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.216343623715147e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5845087786717828e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011932422592384665, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.173638733959198 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.267167866230011, + "mean_feedback_forward_cosine": 0.992410211129622, + "mean_feedback_forward_relative_error": 0.11293627524917776, + "min_feedback_forward_cosine": 0.9649142026901245 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002759207871883248, + "instruction_rms": 0.0027592078716066623, + "raw_apical_rms": 0.11215969665710677, + "teaching_rms": 0.002759207871883248, + "traffic_rms": 0.11212472716686142 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.710946737735867e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5828210412115806e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01089502842617996, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.17161079182624817 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2559974193572998, + "mean_feedback_forward_cosine": 0.993004380572926, + "mean_feedback_forward_relative_error": 0.10842407721687447, + "min_feedback_forward_cosine": 0.9677553176879883 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002808980630273132, + "instruction_rms": 0.0028089806307528176, + "raw_apical_rms": 0.11227211893570996, + "teaching_rms": 0.002808980630273132, + "traffic_rms": 0.11223564725693122 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.98053170822277e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585605006081577e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010594170585223618, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.17664622473716735 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24627509713172913, + "mean_feedback_forward_cosine": 0.9935285253958268, + "mean_feedback_forward_relative_error": 0.10427742729132826, + "min_feedback_forward_cosine": 0.9701375961303711 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002749674188442066, + "instruction_rms": 0.0027496741880878657, + "raw_apical_rms": 0.11261646802200022, + "teaching_rms": 0.002749674188442066, + "traffic_rms": 0.11258166804856107 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 1.0941191419533425e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5943929286696242e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011147741947802742, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16682959839820863 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23575939238071442, + "mean_feedback_forward_cosine": 0.9940485802563754, + "mean_feedback_forward_relative_error": 0.10005330456928774, + "min_feedback_forward_cosine": 0.9726309776306152 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002790564991608982, + "instruction_rms": 0.0027905649915647528, + "raw_apical_rms": 0.11319391358949094, + "teaching_rms": 0.002790564991608982, + "traffic_rms": 0.11315844489251131 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.19809839008667e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5870919806479708e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011355676204249326, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.17255522152900696 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22681856155395508, + "mean_feedback_forward_cosine": 0.9944915576414628, + "mean_feedback_forward_relative_error": 0.0962677910585295, + "min_feedback_forward_cosine": 0.9746420383453369 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027102017553432486, + "instruction_rms": 0.0027102017555373065, + "raw_apical_rms": 0.11342980849805166, + "teaching_rms": 0.0027102017553432486, + "traffic_rms": 0.11339641834719952 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.695326166232917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5856189912252844e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011063832872983907, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16799449390411378 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21755829453468323, + "mean_feedback_forward_cosine": 0.9949110703034835, + "mean_feedback_forward_relative_error": 0.09257623702287673, + "min_feedback_forward_cosine": 0.9766520261764526 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002705806527157039, + "instruction_rms": 0.0027058065268196698, + "raw_apical_rms": 0.11372310299273025, + "teaching_rms": 0.002705806527157039, + "traffic_rms": 0.11368980567945332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.974482064630138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5771032542605507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011658218361989088, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.16816208700180055 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20857210457324982, + "mean_feedback_forward_cosine": 0.9952987324107777, + "mean_feedback_forward_relative_error": 0.0890488951382312, + "min_feedback_forward_cosine": 0.9785271883010864 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002667031936576862, + "instruction_rms": 0.0026670319367579015, + "raw_apical_rms": 0.11417956981020295, + "teaching_rms": 0.002667031936576862, + "traffic_rms": 0.11414735620657027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.931807033401128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5912598547130218e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011491496180510636, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16299987187862397 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19948440790176392, + "mean_feedback_forward_cosine": 0.9956702470779419, + "mean_feedback_forward_relative_error": 0.08551065349443392, + "min_feedback_forward_cosine": 0.9803446531295776 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00270891648821242, + "instruction_rms": 0.002708916488371943, + "raw_apical_rms": 0.1147552486085405, + "teaching_rms": 0.00270891648821242, + "traffic_rms": 0.11472209748369899 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.797106794602769e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.57913181086537e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01154276258741217, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1694073840165138 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1919146180152893, + "mean_feedback_forward_cosine": 0.9959955096244812, + "mean_feedback_forward_relative_error": 0.08224443102424794, + "min_feedback_forward_cosine": 0.9818003177642822 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026757368272720817, + "instruction_rms": 0.0026757368279129025, + "raw_apical_rms": 0.11465049734666755, + "teaching_rms": 0.0026757368272720817, + "traffic_rms": 0.11461804686354728 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.556254561473907e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5777359615018396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012158770954836529, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.16673356837272643 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18463130295276642, + "mean_feedback_forward_cosine": 0.9962979240850969, + "mean_feedback_forward_relative_error": 0.07910033538937569, + "min_feedback_forward_cosine": 0.9831477403640747 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002703043145014737, + "instruction_rms": 0.0027030431453699865, + "raw_apical_rms": 0.11496117112493232, + "teaching_rms": 0.002703043145014737, + "traffic_rms": 0.11492822372276584 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 4.277335108326952e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5839071604137478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012131877471975843, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15652737518787385 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17719042301177979, + "mean_feedback_forward_cosine": 0.9965843894264915, + "mean_feedback_forward_relative_error": 0.07599363960325718, + "min_feedback_forward_cosine": 0.9844739437103271 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002738979411209901, + "instruction_rms": 0.002738979411246894, + "raw_apical_rms": 0.1151913874906174, + "teaching_rms": 0.002738979411209901, + "traffic_rms": 0.11515761051080645 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.2633296832027554e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5900035032935614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012149310516543482, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.1626805915427208 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16977755725383759, + "mean_feedback_forward_cosine": 0.9968520272861827, + "mean_feedback_forward_relative_error": 0.07300305214117873, + "min_feedback_forward_cosine": 0.9857430458068848 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744472035960319, + "instruction_rms": 0.002744472036072214, + "raw_apical_rms": 0.11566503738715905, + "teaching_rms": 0.002744472035960319, + "traffic_rms": 0.11563130908646636 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.2333170685869845e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5863550233155286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011248005839427094, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.16187418454170227 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16306212544441223, + "mean_feedback_forward_cosine": 0.9970953757112676, + "mean_feedback_forward_relative_error": 0.07014588591727343, + "min_feedback_forward_cosine": 0.9868466854095459 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002721329132782141, + "instruction_rms": 0.0027213291330304523, + "raw_apical_rms": 0.1157166554505299, + "teaching_rms": 0.002721329132782141, + "traffic_rms": 0.11568349094031427 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.354844662657342e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5906972708248838e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011900405117934287, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.16319609530448914 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15640316903591156, + "mean_feedback_forward_cosine": 0.9973227284171364, + "mean_feedback_forward_relative_error": 0.06737760383297096, + "min_feedback_forward_cosine": 0.9878937602043152 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002740100736707151, + "instruction_rms": 0.0027401007368791563, + "raw_apical_rms": 0.11630187949384375, + "teaching_rms": 0.002740100736707151, + "traffic_rms": 0.11626842569493483 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.008018888474908e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5893530047018398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01164139687944173, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.16348790993571283 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15050576627254486, + "mean_feedback_forward_cosine": 0.9975215695121071, + "mean_feedback_forward_relative_error": 0.06483579948544502, + "min_feedback_forward_cosine": 0.9887878894805908 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027230809056385096, + "instruction_rms": 0.002723080906629212, + "raw_apical_rms": 0.11652576019709164, + "teaching_rms": 0.0027230809056385096, + "traffic_rms": 0.11649278796198989 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.699724508507643e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5909279265196596e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011933579543743874, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15690901566505433 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1443258672952652, + "mean_feedback_forward_cosine": 0.9977171594446356, + "mean_feedback_forward_relative_error": 0.06225192980332808, + "min_feedback_forward_cosine": 0.9896854758262634 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002751302036198619, + "instruction_rms": 0.0027513020366263107, + "raw_apical_rms": 0.11650580467202318, + "teaching_rms": 0.002751302036198619, + "traffic_rms": 0.11647200766942926 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.0853468991363115e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.596230245969911e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011138478931621383, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15821535417079927 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13850119709968567, + "mean_feedback_forward_cosine": 0.9978924263607372, + "mean_feedback_forward_relative_error": 0.05985670672221617, + "min_feedback_forward_cosine": 0.9904961585998535 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002691741844296364, + "instruction_rms": 0.002691741844050014, + "raw_apical_rms": 0.11697868076335545, + "teaching_rms": 0.002691741844296364, + "traffic_rms": 0.11694645871523873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.329877612008204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5888427527211002e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011764964354133602, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.1524140273809433 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13290835916996002, + "mean_feedback_forward_cosine": 0.9980539560317994, + "mean_feedback_forward_relative_error": 0.05754290831021287, + "min_feedback_forward_cosine": 0.991249144077301 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027153065571034404, + "instruction_rms": 0.0027153065578077104, + "raw_apical_rms": 0.11727591544398375, + "teaching_rms": 0.0027153065571034404, + "traffic_rms": 0.11724330261499231 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.6912933981957394e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.594280566201682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01116302196931127, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.15904617218494416 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12745355069637299, + "mean_feedback_forward_cosine": 0.9982035105878656, + "mean_feedback_forward_relative_error": 0.055316212400794026, + "min_feedback_forward_cosine": 0.991946816444397 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027012344886789726, + "instruction_rms": 0.0027012344883947936, + "raw_apical_rms": 0.11756471256513797, + "teaching_rms": 0.0027012344886789726, + "traffic_rms": 0.11753247528634415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.705789180547981e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5973163856463467e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010616420033518533, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.15437325385570527 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12211593985557556, + "mean_feedback_forward_cosine": 0.9983466310934587, + "mean_feedback_forward_relative_error": 0.05310003813356161, + "min_feedback_forward_cosine": 0.9926053285598755 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002738522249226644, + "instruction_rms": 0.002738522249506231, + "raw_apical_rms": 0.11780735484506301, + "teaching_rms": 0.002738522249226644, + "traffic_rms": 0.11777425955961487 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.298473375134563e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585730881467693e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010981096515611915, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1542772695016861 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11682494729757309, + "mean_feedback_forward_cosine": 0.9984773148189892, + "mean_feedback_forward_relative_error": 0.050998348505659534, + "min_feedback_forward_cosine": 0.9932323694229126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002723280323955817, + "instruction_rms": 0.0027232803231945287, + "raw_apical_rms": 0.11821450858497025, + "teaching_rms": 0.002723280323955817, + "traffic_rms": 0.11818177823843887 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.071095543485171e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.592598595109569e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010903889070222344, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.15452689791202545 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11196722835302353, + "mean_feedback_forward_cosine": 0.9985977162014354, + "mean_feedback_forward_relative_error": 0.04896128528158773, + "min_feedback_forward_cosine": 0.9937810897827148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027691645858947686, + "instruction_rms": 0.002769164586154739, + "raw_apical_rms": 0.11830188999049981, + "teaching_rms": 0.0027691645858947686, + "traffic_rms": 0.1182682797242431 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.037965588096995e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5911204203264162e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011073834683789148, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.15595438538551332 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1075218915939331, + "mean_feedback_forward_cosine": 0.9987039901993492, + "mean_feedback_forward_relative_error": 0.047091127661141485, + "min_feedback_forward_cosine": 0.9942640066146851 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026887574276137256, + "instruction_rms": 0.0026887574268776963, + "raw_apical_rms": 0.11864462754438557, + "teaching_rms": 0.0026887574276137256, + "traffic_rms": 0.11861282294762501 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.13417468539501e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5962443382540246e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010811256691004769, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1518903250002861 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10297679156064987, + "mean_feedback_forward_cosine": 0.9988058079372752, + "mean_feedback_forward_relative_error": 0.04523424465547909, + "min_feedback_forward_cosine": 0.9947372078895569 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002708434850801182, + "instruction_rms": 0.002708434851284277, + "raw_apical_rms": 0.1189415164862091, + "teaching_rms": 0.002708434850801182, + "traffic_rms": 0.11890941686819564 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.562076310638986e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5991038007164013e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012129613038889406, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.15055798265457154 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0987158715724945, + "mean_feedback_forward_cosine": 0.9988990328528664, + "mean_feedback_forward_relative_error": 0.043450768427415326, + "min_feedback_forward_cosine": 0.9951611757278442 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027739074744654074, + "instruction_rms": 0.002773907473822653, + "raw_apical_rms": 0.11920308798426466, + "teaching_rms": 0.0027739074744654074, + "traffic_rms": 0.1191695404635708 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 4.660348551510651e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5974532874985374e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013024560553677775, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15162084616661073 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09467765688896179, + "mean_feedback_forward_cosine": 0.9989832021973349, + "mean_feedback_forward_relative_error": 0.041772408461706205, + "min_feedback_forward_cosine": 0.9955476522445679 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744016569937722, + "instruction_rms": 0.0027440165705264283, + "raw_apical_rms": 0.11948916186934745, + "teaching_rms": 0.002744016569937722, + "traffic_rms": 0.11945641895994157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.182579698725931e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587550128513507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013334009979847088, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14631768272876738 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09055773913860321, + "mean_feedback_forward_cosine": 0.9990655378861861, + "mean_feedback_forward_relative_error": 0.040071847001937305, + "min_feedback_forward_cosine": 0.9959251284599304 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027838001193236555, + "instruction_rms": 0.00278380011909235, + "raw_apical_rms": 0.11960974302303815, + "teaching_rms": 0.0027838001193236555, + "traffic_rms": 0.11957617217582032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.449512651191981e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5969476450535596e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012342750499031286, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.15604819481372834 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08688408881425858, + "mean_feedback_forward_cosine": 0.9991355039856651, + "mean_feedback_forward_relative_error": 0.03855858841402964, + "min_feedback_forward_cosine": 0.9962475299835205 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026812023653437563, + "instruction_rms": 0.002681202365234879, + "raw_apical_rms": 0.11997271730279067, + "teaching_rms": 0.0026812023653437563, + "traffic_rms": 0.11994161209036448 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.720367906609681e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5813674139351247e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0123046264848254, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.15090042344093324 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08323752135038376, + "mean_feedback_forward_cosine": 0.9992031162435359, + "mean_feedback_forward_relative_error": 0.037040479024025526, + "min_feedback_forward_cosine": 0.9965547919273376 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027337911424862105, + "instruction_rms": 0.0027337911422577526, + "raw_apical_rms": 0.12036681863764069, + "teaching_rms": 0.0027337911424862105, + "traffic_rms": 0.12033446189964718 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.627863223864551e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5885021960486138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011901115210563454, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.1471392963409424 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07995020598173141, + "mean_feedback_forward_cosine": 0.9992632139812816, + "mean_feedback_forward_relative_error": 0.035634634626859965, + "min_feedback_forward_cosine": 0.9968221783638 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026814769482426982, + "instruction_rms": 0.0026814769479257057, + "raw_apical_rms": 0.1207011112405806, + "teaching_rms": 0.0026814769482426982, + "traffic_rms": 0.1206701136312538 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.985449374496966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5908765378047952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012255721957310466, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14622336290836335 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0767374187707901, + "mean_feedback_forward_cosine": 0.9993208495053378, + "mean_feedback_forward_relative_error": 0.03422663713043386, + "min_feedback_forward_cosine": 0.9970715045928955 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027147853820148437, + "instruction_rms": 0.002714785382084618, + "raw_apical_rms": 0.12108754828411754, + "teaching_rms": 0.0027147853820148437, + "traffic_rms": 0.12105586021879858 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.575009336984294e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5841440496634433e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012909678366106835, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.1493096795129776 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07371226698160172, + "mean_feedback_forward_cosine": 0.9993749564344233, + "mean_feedback_forward_relative_error": 0.03284511249512434, + "min_feedback_forward_cosine": 0.9972981810569763 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027273905069123763, + "instruction_rms": 0.002727390506936166, + "raw_apical_rms": 0.12114865221405499, + "teaching_rms": 0.0027273905069123763, + "traffic_rms": 0.12111668693906504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.197735302748697e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5888557645784307e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.013382915168470417, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1504336270260811 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07094317674636841, + "mean_feedback_forward_cosine": 0.9994208921085704, + "mean_feedback_forward_relative_error": 0.03163022338836031, + "min_feedback_forward_cosine": 0.9974968433380127 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002657818008155071, + "instruction_rms": 0.002657818007555464, + "raw_apical_rms": 0.12141300245350525, + "teaching_rms": 0.002657818008155071, + "traffic_rms": 0.12138264261555164 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.7002676356887605e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.582622917107439e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.012760869669393275, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1436343447494507 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091907411813736, + "mean_feedback_forward_cosine": 0.9994212854992259, + "mean_feedback_forward_relative_error": 0.031618769382211295, + "min_feedback_forward_cosine": 0.9974988102912903 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019316313622018838, + "instruction_rms": 0.0019316313624522811, + "raw_apical_rms": 0.1215034387457816, + "teaching_rms": 0.0019316313622018838, + "traffic_rms": 0.1214870434907806 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.989663520722388e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.581989970804751e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01150312864546241, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.0774578609919548 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07090752571821213, + "mean_feedback_forward_cosine": 0.9994214816526933, + "mean_feedback_forward_relative_error": 0.03161275628954172, + "min_feedback_forward_cosine": 0.9974995851516724 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016164187386532483, + "instruction_rms": 0.0016164187379367318, + "raw_apical_rms": 0.12143790922360295, + "teaching_rms": 0.0016164187386532483, + "traffic_rms": 0.12142615277399116 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.946496000637126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5822570904228345e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011240200926792239, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.05026715766429901 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708957388997078, + "mean_feedback_forward_cosine": 0.9994216778061606, + "mean_feedback_forward_relative_error": 0.03160706502808766, + "min_feedback_forward_cosine": 0.9975005984306335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015655895145317818, + "instruction_rms": 0.0015655895143415635, + "raw_apical_rms": 0.12146518803248786, + "teaching_rms": 0.0015655895145317818, + "traffic_rms": 0.12145392412925508 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.6390728531141576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.588699782526619e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011209815259896283, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.04198218698382378 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088350504636765, + "mean_feedback_forward_cosine": 0.9994218674573031, + "mean_feedback_forward_relative_error": 0.03160143613137982, + "min_feedback_forward_cosine": 0.9975016117095947 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015589449444097573, + "instruction_rms": 0.0015589449438359359, + "raw_apical_rms": 0.12146266777793177, + "teaching_rms": 0.0015589449444097573, + "traffic_rms": 0.12145131129030932 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.02598717591718e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5921731102529307e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011442721253294508, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.038764991085529325 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087132334709167, + "mean_feedback_forward_cosine": 0.9994220516898416, + "mean_feedback_forward_relative_error": 0.031596079951321535, + "min_feedback_forward_cosine": 0.9975024461746216 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014946659889936184, + "instruction_rms": 0.0014946659893532994, + "raw_apical_rms": 0.12136492652307224, + "teaching_rms": 0.0014946659889936184, + "traffic_rms": 0.12135441252202359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.8556929022634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5925731214650785e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011466905693004414, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03346785994887352 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07085894793272018, + "mean_feedback_forward_cosine": 0.9994222315874967, + "mean_feedback_forward_relative_error": 0.03159078823571855, + "min_feedback_forward_cosine": 0.997503399848938 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014659991795054211, + "instruction_rms": 0.0014659991795446482, + "raw_apical_rms": 0.12133958582295583, + "teaching_rms": 0.0014659991795054211, + "traffic_rms": 0.12132920406922273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.191875456806883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5812456358247876e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0113622764933214, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.030965674679875373 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07084622234106064, + "mean_feedback_forward_cosine": 0.9994224266572432, + "mean_feedback_forward_relative_error": 0.03158551456237381, + "min_feedback_forward_cosine": 0.9975043535232544 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014171350812077195, + "instruction_rms": 0.0014171350809680833, + "raw_apical_rms": 0.12126953352127469, + "teaching_rms": 0.0014171350812077195, + "traffic_rms": 0.12125969491249729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.95457637536856e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5867696865059138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011323451646696788, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.026601712737083436 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07083134353160858, + "mean_feedback_forward_cosine": 0.9994226228107106, + "mean_feedback_forward_relative_error": 0.03158002750609409, + "min_feedback_forward_cosine": 0.9975054264068604 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014714192387132573, + "instruction_rms": 0.0014714192389506454, + "raw_apical_rms": 0.12123142209990041, + "teaching_rms": 0.0014714192387132573, + "traffic_rms": 0.12122074245074443 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.581669254046574e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5877256849609277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01136744383785869, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.02691594778686762 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07081860303878784, + "mean_feedback_forward_cosine": 0.9994228027083657, + "mean_feedback_forward_relative_error": 0.03157492659308694, + "min_feedback_forward_cosine": 0.997506320476532 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014073182534575094, + "instruction_rms": 0.001407318253545341, + "raw_apical_rms": 0.12119996807338988, + "teaching_rms": 0.0014073182534575094, + "traffic_rms": 0.12119018610667305 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.822034492870444e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5886854829000422e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011277373316037512, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.02352189971268177 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07080463320016861, + "mean_feedback_forward_cosine": 0.9994230021129955, + "mean_feedback_forward_relative_error": 0.03156958261335438, + "min_feedback_forward_cosine": 0.9975072741508484 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014099075998892502, + "instruction_rms": 0.0014099075998816066, + "raw_apical_rms": 0.12112123411919175, + "teaching_rms": 0.0014099075998892502, + "traffic_rms": 0.12111119762902137 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.605427588761813e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5821945758457385e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011302285868101745, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.022769682832360267 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078919559717178, + "mean_feedback_forward_cosine": 0.999423191764138, + "mean_feedback_forward_relative_error": 0.031564079589125785, + "min_feedback_forward_cosine": 0.9975082874298096 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013808268351238621, + "instruction_rms": 0.0013808268347054852, + "raw_apical_rms": 0.12106233088775874, + "teaching_rms": 0.0013808268351238621, + "traffic_rms": 0.12105244650249325 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.174393973878978e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5891857941082134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011206880855694566, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.02159861841365695 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077489793300629, + "mean_feedback_forward_cosine": 0.9994233857501637, + "mean_feedback_forward_relative_error": 0.03155893870544704, + "min_feedback_forward_cosine": 0.9975095391273499 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013028560768730934, + "instruction_rms": 0.0013028560771809524, + "raw_apical_rms": 0.12108529178479369, + "teaching_rms": 0.0013028560768730934, + "traffic_rms": 0.12107654301746641 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.978675060054229e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583276785733679e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011263281398540473, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.01856324349567294 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707588866353035, + "mean_feedback_forward_cosine": 0.9994235873222351, + "mean_feedback_forward_relative_error": 0.031553159874271264, + "min_feedback_forward_cosine": 0.9975104928016663 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014013425898332828, + "instruction_rms": 0.0014013425899619726, + "raw_apical_rms": 0.121081403301387, + "teaching_rms": 0.0014013425898332828, + "traffic_rms": 0.12107097980157637 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.024830412343363e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5882152459600006e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011302325388719961, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.019349969775676726 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07074145972728729, + "mean_feedback_forward_cosine": 0.9994238051501187, + "mean_feedback_forward_relative_error": 0.031547342105345294, + "min_feedback_forward_cosine": 0.9975118637084961 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014052185171918158, + "instruction_rms": 0.0014052185175645725, + "raw_apical_rms": 0.12102726742081532, + "teaching_rms": 0.0014052185171918158, + "traffic_rms": 0.12101690966033238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.61105090671299e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591355540991409e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011147686507963026, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.019647072715163232 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07072681933641434, + "mean_feedback_forward_cosine": 0.9994239980524237, + "mean_feedback_forward_relative_error": 0.031542143395001235, + "min_feedback_forward_cosine": 0.9975128173828125 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001276663943904826, + "instruction_rms": 0.0012766639440084527, + "raw_apical_rms": 0.12103132502320636, + "teaching_rms": 0.001276663943904826, + "traffic_rms": 0.12102249971396088 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.019663345568915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5902621425555367e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011040123102951115, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.01670920530408621 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07071109116077423, + "mean_feedback_forward_cosine": 0.999424211545424, + "mean_feedback_forward_relative_error": 0.031536509028889916, + "min_feedback_forward_cosine": 0.9975141286849976 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001341280220595386, + "instruction_rms": 0.0013412802208687513, + "raw_apical_rms": 0.12107088186603233, + "teaching_rms": 0.001341280220595386, + "traffic_rms": 0.12106134885199238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.646929324143457e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5871619346364454e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01112268940188316, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.017786935017704963 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07069508731365204, + "mean_feedback_forward_cosine": 0.9994244282895869, + "mean_feedback_forward_relative_error": 0.031530873756855725, + "min_feedback_forward_cosine": 0.9975152015686035 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001296360008357897, + "instruction_rms": 0.0012963600082679932, + "raw_apical_rms": 0.1210064433982024, + "teaching_rms": 0.001296360008357897, + "traffic_rms": 0.12099698297409692 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.4556799966576364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5842384237502345e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011258379301681162, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.01594989131093025 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07067932188510895, + "mean_feedback_forward_cosine": 0.9994246320290999, + "mean_feedback_forward_relative_error": 0.031525330025364055, + "min_feedback_forward_cosine": 0.9975162744522095 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012886175056441664, + "instruction_rms": 0.0012886175053394382, + "raw_apical_rms": 0.12096846561161184, + "teaching_rms": 0.0012886175056441664, + "traffic_rms": 0.12095910561557505 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.389987644046414e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5857442045117193e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01129545211510929, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.01590382546454668 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07066480815410614, + "mean_feedback_forward_cosine": 0.9994248108430343, + "mean_feedback_forward_relative_error": 0.03152040261775255, + "min_feedback_forward_cosine": 0.9975171685218811 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001176883061617576, + "instruction_rms": 0.0011768830620613702, + "raw_apical_rms": 0.12093138647214617, + "teaching_rms": 0.001176883061617576, + "traffic_rms": 0.12092358137503635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.7555919141805134e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583027312864346e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011270806154516505, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.013526519335210324 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07064840197563171, + "mean_feedback_forward_cosine": 0.9994250243360346, + "mean_feedback_forward_relative_error": 0.03151499366557056, + "min_feedback_forward_cosine": 0.9975184202194214 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012067306893248804, + "instruction_rms": 0.0012067306892235943, + "raw_apical_rms": 0.12093184660216626, + "teaching_rms": 0.0012067306893248804, + "traffic_rms": 0.12092325626158301 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.8383195045717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5873000785880004e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011278642957340152, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.013874591251760722 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07063398510217667, + "mean_feedback_forward_cosine": 0.9994252139871771, + "mean_feedback_forward_relative_error": 0.03151010434058579, + "min_feedback_forward_cosine": 0.9975194931030273 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011215831796233238, + "instruction_rms": 0.0011215831800664925, + "raw_apical_rms": 0.12092683518032567, + "teaching_rms": 0.0011215831796233238, + "traffic_rms": 0.12091950006035208 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 9.060403982630305e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5902660829879754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01131065804714938, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.012541126489341258 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07061872631311417, + "mean_feedback_forward_cosine": 0.9994253873825073, + "mean_feedback_forward_relative_error": 0.031505210553719236, + "min_feedback_forward_cosine": 0.9975205063819885 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011016172111567837, + "instruction_rms": 0.001101617211390144, + "raw_apical_rms": 0.12088319494619222, + "teaching_rms": 0.0011016172111567837, + "traffic_rms": 0.12087577071212233 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.362741172300957e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5800089468773435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011174570979947764, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.012012996182143689 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07060375809669495, + "mean_feedback_forward_cosine": 0.9994255900382996, + "mean_feedback_forward_relative_error": 0.03150028646500273, + "min_feedback_forward_cosine": 0.9975217580795288 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010720577735974454, + "instruction_rms": 0.0010720577734946305, + "raw_apical_rms": 0.12087417070598201, + "teaching_rms": 0.0010720577735974454, + "traffic_rms": 0.12086698554962963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.824389859299117e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5860020862468105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011193659388315395, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.011400184899494051 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07058724015951157, + "mean_feedback_forward_cosine": 0.9994257872754877, + "mean_feedback_forward_relative_error": 0.03149497618221424, + "min_feedback_forward_cosine": 0.99752277135849 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011329482021261426, + "instruction_rms": 0.0011329482019174382, + "raw_apical_rms": 0.12087773359409643, + "teaching_rms": 0.0011329482021261426, + "traffic_rms": 0.12086960561142271 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 9.533328437782984e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587363652648783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011133339256990647, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.01149278408486396 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07057030498981476, + "mean_feedback_forward_cosine": 0.9994260018522089, + "mean_feedback_forward_relative_error": 0.03148964259942824, + "min_feedback_forward_cosine": 0.9975239634513855 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011235810648741928, + "instruction_rms": 0.001123581065121974, + "raw_apical_rms": 0.12088743305799948, + "teaching_rms": 0.0011235810648741928, + "traffic_rms": 0.12087967859722216 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.870171548849612e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5838497823317857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011213246111817024, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.011122201923280955 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07055456191301346, + "mean_feedback_forward_cosine": 0.9994261828335849, + "mean_feedback_forward_relative_error": 0.031484583401205864, + "min_feedback_forward_cosine": 0.9975250959396362 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010577879550685485, + "instruction_rms": 0.0010577879551434022, + "raw_apical_rms": 0.12091154495210744, + "teaching_rms": 0.0010577879550685485, + "traffic_rms": 0.12090424319941545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.408785679923312e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5814129169934688e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011142524557366121, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.010322054066061974 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07053574919700623, + "mean_feedback_forward_cosine": 0.9994264060800726, + "mean_feedback_forward_relative_error": 0.03147879388522018, + "min_feedback_forward_cosine": 0.9975264072418213 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001133892089154218, + "instruction_rms": 0.001133892089078077, + "raw_apical_rms": 0.12093082090073219, + "teaching_rms": 0.001133892089154218, + "traffic_rms": 0.12092282742761554 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.354558562155944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5912700142957998e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011092551928234894, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.010840231428295374 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07051771879196167, + "mean_feedback_forward_cosine": 0.9994266314940019, + "mean_feedback_forward_relative_error": 0.03147308212620291, + "min_feedback_forward_cosine": 0.9975277185440063 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011080609356600107, + "instruction_rms": 0.001108060936019147, + "raw_apical_rms": 0.12088387004409894, + "teaching_rms": 0.0011080609356600107, + "traffic_rms": 0.12087612855137313 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.18298727772526e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5840815336388044e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01117625895196831, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.010595754879266023 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07050331681966782, + "mean_feedback_forward_cosine": 0.999426823312586, + "mean_feedback_forward_relative_error": 0.03146810530769554, + "min_feedback_forward_cosine": 0.9975285530090332 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010294044538193266, + "instruction_rms": 0.0010294044539697284, + "raw_apical_rms": 0.12077501785067396, + "teaching_rms": 0.0010294044538193266, + "traffic_rms": 0.12076823623982696 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.054079193518191e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5853927420484418e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011366229509138544, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.0090453971606493 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07048619538545609, + "mean_feedback_forward_cosine": 0.9994270465590737, + "mean_feedback_forward_relative_error": 0.0314625168134543, + "min_feedback_forward_cosine": 0.9975297451019287 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010563640294215934, + "instruction_rms": 0.0010563640296493582, + "raw_apical_rms": 0.12083677535731305, + "teaching_rms": 0.0010563640294215934, + "traffic_rms": 0.12082958511468528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.440790656370154e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5845864610262134e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01114550376388394, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.009695652887299656 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07046898454427719, + "mean_feedback_forward_cosine": 0.9994272698055614, + "mean_feedback_forward_relative_error": 0.03145684924145991, + "min_feedback_forward_cosine": 0.9975310564041138 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010952395358122472, + "instruction_rms": 0.001095239535493576, + "raw_apical_rms": 0.12084477594639009, + "teaching_rms": 0.0010952395358122472, + "traffic_rms": 0.12083691550699845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.913351313642124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5865155531216587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011374885523999454, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.009708992003947496 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07045096904039383, + "mean_feedback_forward_cosine": 0.999427464875308, + "mean_feedback_forward_relative_error": 0.03145115019922907, + "min_feedback_forward_cosine": 0.997532308101654 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010348551966495592, + "instruction_rms": 0.0010348551969669674, + "raw_apical_rms": 0.12083036618657504, + "teaching_rms": 0.0010348551966495592, + "traffic_rms": 0.12082286248496703 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.677312486795927e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5867192692839256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011345734802015235, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.009417185395061969 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07043527811765671, + "mean_feedback_forward_cosine": 0.9994276751171459, + "mean_feedback_forward_relative_error": 0.031446041124449535, + "min_feedback_forward_cosine": 0.9975335597991943 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009991465308276924, + "instruction_rms": 0.000999146530843497, + "raw_apical_rms": 0.12079991080667568, + "teaching_rms": 0.0009991465308276924, + "traffic_rms": 0.12079349202199281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.6335728970680066e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5813381485782494e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01121630788174555, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.008770374926365912 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07041771709918976, + "mean_feedback_forward_cosine": 0.9994278896938671, + "mean_feedback_forward_relative_error": 0.031440461105243726, + "min_feedback_forward_cosine": 0.9975347518920898 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010425932026417073, + "instruction_rms": 0.0010425932027485045, + "raw_apical_rms": 0.12075980471417198, + "teaching_rms": 0.0010425932026417073, + "traffic_rms": 0.12075269167967503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.208813540022675e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5942978381964335e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011360938807897156, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.009208371835425496 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0704011470079422, + "mean_feedback_forward_cosine": 0.9994281096891924, + "mean_feedback_forward_relative_error": 0.03143500339409167, + "min_feedback_forward_cosine": 0.9975358247756958 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010104558630069274, + "instruction_rms": 0.0010104558629655414, + "raw_apical_rms": 0.12077581042787909, + "teaching_rms": 0.0010104558630069274, + "traffic_rms": 0.12076903893089148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.122106555039863e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5858008022959945e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011201823207306057, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.00852915184430778 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07038401812314987, + "mean_feedback_forward_cosine": 0.9994283383542841, + "mean_feedback_forward_relative_error": 0.03142943127419461, + "min_feedback_forward_cosine": 0.9975371360778809 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009940564759869151, + "instruction_rms": 0.000994056476289328, + "raw_apical_rms": 0.12077183782974174, + "teaching_rms": 0.0009940564759869151, + "traffic_rms": 0.1207648402222815 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.744818392689922e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583754656607176e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010985859320546711, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.008393654064536094 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07036565989255905, + "mean_feedback_forward_cosine": 0.9994285529310053, + "mean_feedback_forward_relative_error": 0.03142367072234099, + "min_feedback_forward_cosine": 0.9975384473800659 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010190824464492904, + "instruction_rms": 0.0010190824463925693, + "raw_apical_rms": 0.12075393140369606, + "teaching_rms": 0.0010190824464492904, + "traffic_rms": 0.12074701113834552 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.937292613111822e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.588657046772156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010946906221500706, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.007965772947594523 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07034841179847717, + "mean_feedback_forward_cosine": 0.9994287696751681, + "mean_feedback_forward_relative_error": 0.03141821456903761, + "min_feedback_forward_cosine": 0.9975396394729614 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000934028921516271, + "instruction_rms": 0.0009340289212301821, + "raw_apical_rms": 0.1207637690767968, + "teaching_rms": 0.000934028921516271, + "traffic_rms": 0.12075757479166307 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.48463327479476e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5861000153503188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011057929100270425, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.007507783609703183 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0703318864107132, + "mean_feedback_forward_cosine": 0.9994289701635187, + "mean_feedback_forward_relative_error": 0.03141317000104622, + "min_feedback_forward_cosine": 0.9975408911705017 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009013686103250061, + "instruction_rms": 0.0009013686101781583, + "raw_apical_rms": 0.12071825369750805, + "teaching_rms": 0.0009013686103250061, + "traffic_rms": 0.12071265689185987 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.831189836020712e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5880163588553643e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011220030254864954, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.00748919204108417 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07031499594449997, + "mean_feedback_forward_cosine": 0.999429195577448, + "mean_feedback_forward_relative_error": 0.031407627336342225, + "min_feedback_forward_cosine": 0.9975419640541077 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009761039731833493, + "instruction_rms": 0.000976103973289089, + "raw_apical_rms": 0.12070630531065162, + "teaching_rms": 0.0009761039731833493, + "traffic_rms": 0.12069970974373335 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.592161400613804e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589194410847404e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011198168016916228, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.00756540226444602 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07029535621404648, + "mean_feedback_forward_cosine": 0.9994294220750982, + "mean_feedback_forward_relative_error": 0.03140136955136603, + "min_feedback_forward_cosine": 0.9975433349609375 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009887591687614944, + "instruction_rms": 0.0009887591692578233, + "raw_apical_rms": 0.12071824863694143, + "teaching_rms": 0.0009887591687614944, + "traffic_rms": 0.12071114896001232 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.8462767071887356e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5848191904646026e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01122735820319111, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.00782873647440225 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07027541846036911, + "mean_feedback_forward_cosine": 0.9994296734983271, + "mean_feedback_forward_relative_error": 0.031395242651077834, + "min_feedback_forward_cosine": 0.9975447058677673 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009748575675087432, + "instruction_rms": 0.0009748575673268936, + "raw_apical_rms": 0.12076131454864471, + "teaching_rms": 0.0009748575675087432, + "traffic_rms": 0.12075429179352912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.705413906933245e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5866352186913332e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011196456742615693, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.00777764855876565 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07025779038667679, + "mean_feedback_forward_cosine": 0.999429895661094, + "mean_feedback_forward_relative_error": 0.031389443906532095, + "min_feedback_forward_cosine": 0.9975459575653076 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009434352706285746, + "instruction_rms": 0.0009434352704947054, + "raw_apical_rms": 0.12075522034551249, + "teaching_rms": 0.0009434352706285746, + "traffic_rms": 0.12074866983924204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.134634844556786e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5863888524493617e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011169083543723663, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.007418532796762883 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07023657858371735, + "mean_feedback_forward_cosine": 0.999430175261064, + "mean_feedback_forward_relative_error": 0.03138248405673287, + "min_feedback_forward_cosine": 0.9975473880767822 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010115940486411486, + "instruction_rms": 0.0010115940489804026, + "raw_apical_rms": 0.12074958993793988, + "teaching_rms": 0.0010115940486411486, + "traffic_rms": 0.12074189484607889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.299540444842933e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587309086780064e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011167800983200398, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.007985664666891098 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07021334022283554, + "mean_feedback_forward_cosine": 0.9994304635308006, + "mean_feedback_forward_relative_error": 0.03137529146095568, + "min_feedback_forward_cosine": 0.9975489974021912 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010090336205429283, + "instruction_rms": 0.0010090336205609723, + "raw_apical_rms": 0.12074680283928257, + "teaching_rms": 0.0010090336205429283, + "traffic_rms": 0.12073941123422713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.083522035600254e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.585378086634112e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011277140639654279, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.007911784377619625 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07019465416669846, + "mean_feedback_forward_cosine": 0.9994306911121715, + "mean_feedback_forward_relative_error": 0.03136926259506832, + "min_feedback_forward_cosine": 0.9975502490997314 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009662888563246104, + "instruction_rms": 0.000966288856044449, + "raw_apical_rms": 0.12073499670172054, + "teaching_rms": 0.0009662888563246104, + "traffic_rms": 0.12072837977571471 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.496089438063791e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587221212477568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011256401771457725, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.006909314128682017 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07017727941274643, + "mean_feedback_forward_cosine": 0.9994309078563344, + "mean_feedback_forward_relative_error": 0.03136391824280674, + "min_feedback_forward_cosine": 0.9975515604019165 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000869398138851607, + "instruction_rms": 0.0008693981386695274, + "raw_apical_rms": 0.12066423618798937, + "teaching_rms": 0.000869398138851607, + "traffic_rms": 0.12065867949557355 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.0784638716550035e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5856327820394464e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011204988740657578, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.006455450179949403 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07015566527843475, + "mean_feedback_forward_cosine": 0.9994311701167714, + "mean_feedback_forward_relative_error": 0.03135735793039203, + "min_feedback_forward_cosine": 0.9975528717041016 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009108328233069865, + "instruction_rms": 0.0009108328232173712, + "raw_apical_rms": 0.12064820004474244, + "teaching_rms": 0.0009108328233069865, + "traffic_rms": 0.1206417547556839 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.499374194390839e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.588504741949133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011270251572767884, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.00725371184438467 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07013462483882904, + "mean_feedback_forward_cosine": 0.9994314182888377, + "mean_feedback_forward_relative_error": 0.03135055799714544, + "min_feedback_forward_cosine": 0.9975544214248657 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009929308943153806, + "instruction_rms": 0.0009929308941822965, + "raw_apical_rms": 0.12067018188186542, + "teaching_rms": 0.0009929308943153806, + "traffic_rms": 0.12066312890556309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.568165472093824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587661712584924e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01110846032069937, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.007605920844562352 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011668384075165, + "mean_feedback_forward_cosine": 0.9994316447864879, + "mean_feedback_forward_relative_error": 0.03134507797658444, + "min_feedback_forward_cosine": 0.9975556135177612 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008278443350944605, + "instruction_rms": 0.000827844334770908, + "raw_apical_rms": 0.12063417210462912, + "teaching_rms": 0.0008278443350944605, + "traffic_rms": 0.12062883016818716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.079902448163921e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5861055090235406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011049545253435487, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.006175522373914719 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011643052101135, + "mean_feedback_forward_cosine": 0.9994316523725336, + "mean_feedback_forward_relative_error": 0.03134498400613665, + "min_feedback_forward_cosine": 0.9975556135177612 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008630216981704413, + "instruction_rms": 0.0008630216980826758, + "raw_apical_rms": 0.12061825903365644, + "teaching_rms": 0.0008630216981704413, + "traffic_rms": 0.12061274661674656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.687106963137012e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5897691840843058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011071510585886186, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.005969180923551321 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011627405881882, + "mean_feedback_forward_cosine": 0.9994316437027672, + "mean_feedback_forward_relative_error": 0.03134491436014121, + "min_feedback_forward_cosine": 0.9975556135177612 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007460985699154562, + "instruction_rms": 0.0007460985697244131, + "raw_apical_rms": 0.12060988009720111, + "teaching_rms": 0.0007460985699154562, + "traffic_rms": 0.12060513364013943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.912668399332688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.597947370401401e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01094981160962626, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.005264219559244812 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011613249778748, + "mean_feedback_forward_cosine": 0.9994316567074168, + "mean_feedback_forward_relative_error": 0.03134484869343313, + "min_feedback_forward_cosine": 0.997555673122406 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007420030853294929, + "instruction_rms": 0.0007420030853725111, + "raw_apical_rms": 0.12061358630239877, + "teaching_rms": 0.0007420030853294929, + "traffic_rms": 0.120609119304036 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.203235069004222e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5854405454563713e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01099086252175615, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.0050315053797513245 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011598348617554, + "mean_feedback_forward_cosine": 0.9994316469539296, + "mean_feedback_forward_relative_error": 0.031344785854559055, + "min_feedback_forward_cosine": 0.997555673122406 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007039289267185821, + "instruction_rms": 0.0007039289268821719, + "raw_apical_rms": 0.1206168929630992, + "teaching_rms": 0.0007039289267185821, + "traffic_rms": 0.12061270980422223 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.765618737996192e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591492133799338e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011034298548478259, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004824321654289961 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011587917804718, + "mean_feedback_forward_cosine": 0.9994316610423002, + "mean_feedback_forward_relative_error": 0.0313447261906483, + "min_feedback_forward_cosine": 0.9975557327270508 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006376567373198125, + "instruction_rms": 0.000637656737203598, + "raw_apical_rms": 0.12061587949785221, + "teaching_rms": 0.0006376567373198125, + "traffic_rms": 0.12061244154477092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 7.224610953926458e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5889644525274507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010960997161645742, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.004377018192149699 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011575996875763, + "mean_feedback_forward_cosine": 0.9994316534562544, + "mean_feedback_forward_relative_error": 0.03134466705166481, + "min_feedback_forward_cosine": 0.9975557923316956 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000684842550326603, + "instruction_rms": 0.0006848425504731417, + "raw_apical_rms": 0.12061338656526976, + "teaching_rms": 0.000684842550326603, + "traffic_rms": 0.12060958829388432 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.251095641118809e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5865729027490042e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010901957009038442, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.004573622027058155 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011562585830688, + "mean_feedback_forward_cosine": 0.9994316567074168, + "mean_feedback_forward_relative_error": 0.03134460887787017, + "min_feedback_forward_cosine": 0.9975557327270508 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006325919894719429, + "instruction_rms": 0.0006325919895634825, + "raw_apical_rms": 0.1206109873693738, + "teaching_rms": 0.0006325919894719429, + "traffic_rms": 0.12060712511088602 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.076192787280888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5857873653474568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01094310001443777, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.004217614284660667 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011548429727554, + "mean_feedback_forward_cosine": 0.9994316718795083, + "mean_feedback_forward_relative_error": 0.03134454900229519, + "min_feedback_forward_cosine": 0.9975557327270508 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006650859749439857, + "instruction_rms": 0.0006650859747830863, + "raw_apical_rms": 0.12061356531744602, + "teaching_rms": 0.0006650859749439857, + "traffic_rms": 0.12060975090771466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.470630691841507e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5836846972464166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010997232618201456, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.004617708307979629 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0701153501868248, + "mean_feedback_forward_cosine": 0.9994316588748585, + "mean_feedback_forward_relative_error": 0.03134448761967096, + "min_feedback_forward_cosine": 0.9975557327270508 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006748786243155888, + "instruction_rms": 0.0006748786241336835, + "raw_apical_rms": 0.12061284951864801, + "teaching_rms": 0.0006748786243155888, + "traffic_rms": 0.12060873147569375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.900902059849614e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5831064033897414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010893336485269103, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.004377266821749508 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011522352695465, + "mean_feedback_forward_cosine": 0.9994316642934625, + "mean_feedback_forward_relative_error": 0.031344430292533206, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006392136106285652, + "instruction_rms": 0.000639213610542817, + "raw_apical_rms": 0.12061903297586386, + "teaching_rms": 0.0006392136106285652, + "traffic_rms": 0.12061574652084903 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.732165320182503e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5912027100017034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01093279206987273, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.004207859783023596 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0701151117682457, + "mean_feedback_forward_cosine": 0.9994316556236961, + "mean_feedback_forward_relative_error": 0.0313443701798943, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007046983521651161, + "instruction_rms": 0.0007046983520976156, + "raw_apical_rms": 0.12062403666530974, + "teaching_rms": 0.0007046983521651161, + "traffic_rms": 0.1206198430945124 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.373028410029292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.581310126650428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011110784954911633, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.004733310500979424 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011499255895615, + "mean_feedback_forward_cosine": 0.9994316545399753, + "mean_feedback_forward_relative_error": 0.031344310473650695, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006167286458733131, + "instruction_rms": 0.0006167286462098559, + "raw_apical_rms": 0.1206250868131751, + "teaching_rms": 0.0006167286458733131, + "traffic_rms": 0.12062156704827183 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.828225451727121e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589624689138684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010973290234301863, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.00407535614375025 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070114865899086, + "mean_feedback_forward_cosine": 0.9994316632097417, + "mean_feedback_forward_relative_error": 0.03134424732998013, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006839478143122504, + "instruction_rms": 0.0006839478143399428, + "raw_apical_rms": 0.12062466054900933, + "teaching_rms": 0.0006839478143122504, + "traffic_rms": 0.1206205114262598 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.041233291398385e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5877480478490816e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010878825542593694, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.004494556574672461 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011474668979645, + "mean_feedback_forward_cosine": 0.9994316545399753, + "mean_feedback_forward_relative_error": 0.03134418775073507, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000642523662489967, + "instruction_rms": 0.0006425236625722152, + "raw_apical_rms": 0.1206194658352656, + "teaching_rms": 0.000642523662489967, + "traffic_rms": 0.12061568030619436 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.513977896498545e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.586553023758566e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01101916135820035, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.0043582565813884136 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0701146274805069, + "mean_feedback_forward_cosine": 0.9994316740469499, + "mean_feedback_forward_relative_error": 0.03134413026273251, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006384842519838402, + "instruction_rms": 0.0006384842517300336, + "raw_apical_rms": 0.12062253839769596, + "teaching_rms": 0.0006384842519838402, + "traffic_rms": 0.12061878561847109 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 7.249101940942637e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587119106870252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011001107989414591, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.004121162942238152 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011452317237854, + "mean_feedback_forward_cosine": 0.9994316642934625, + "mean_feedback_forward_relative_error": 0.03134407850659706, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005819218577714439, + "instruction_rms": 0.0005819218577712398, + "raw_apical_rms": 0.12062009257400849, + "teaching_rms": 0.0005819218577714439, + "traffic_rms": 0.12061685934381829 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.8667815494573006e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5946013788187038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01093304578535944, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.00393196282312274 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011440396308899, + "mean_feedback_forward_cosine": 0.9994316762143916, + "mean_feedback_forward_relative_error": 0.03134402310983701, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006128534915819116, + "instruction_rms": 0.0006128534914259528, + "raw_apical_rms": 0.12062387358278188, + "teaching_rms": 0.0006128534915819116, + "traffic_rms": 0.12062063400646014 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.95615653967252e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5895688744278397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011057271258596586, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.004055446383729577 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011427730321884, + "mean_feedback_forward_cosine": 0.9994316816329956, + "mean_feedback_forward_relative_error": 0.03134396695108576, + "min_feedback_forward_cosine": 0.9975557923316956 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006219340893221111, + "instruction_rms": 0.000621934089364157, + "raw_apical_rms": 0.12062265113733023, + "teaching_rms": 0.0006219340893221111, + "traffic_rms": 0.12061944854358539 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.821016995672835e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5857311940361583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011033229812654203, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.004223021179288626 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011416554450989, + "mean_feedback_forward_cosine": 0.9994316772981123, + "mean_feedback_forward_relative_error": 0.03134391290051016, + "min_feedback_forward_cosine": 0.9975559115409851 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006069107759692488, + "instruction_rms": 0.0006069107760468724, + "raw_apical_rms": 0.12062285906868363, + "teaching_rms": 0.0006069107759692488, + "traffic_rms": 0.12061966378384595 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.111820738397e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5878056968063384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01098561561277745, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.0039922459670901295 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011407613754272, + "mean_feedback_forward_cosine": 0.9994316772981123, + "mean_feedback_forward_relative_error": 0.031343862482092595, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000526750621256931, + "instruction_rms": 0.000526750621266273, + "raw_apical_rms": 0.12062209701356255, + "teaching_rms": 0.000526750621256931, + "traffic_rms": 0.12061951098282782 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.58036248321514e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5878585251587876e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011095856784359296, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0035031908175349236 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011396437883377, + "mean_feedback_forward_cosine": 0.9994316816329956, + "mean_feedback_forward_relative_error": 0.031343811852010815, + "min_feedback_forward_cosine": 0.9975559115409851 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00058814731639813, + "instruction_rms": 0.000588147316064284, + "raw_apical_rms": 0.12062577858598894, + "teaching_rms": 0.00058814731639813, + "traffic_rms": 0.12062276643304598 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.418833547490067e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.586938087678919e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011049156085271485, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.0037379285618662833 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011387497186661, + "mean_feedback_forward_cosine": 0.9994316838004372, + "mean_feedback_forward_relative_error": 0.03134375771676952, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005767004225555261, + "instruction_rms": 0.0005767004224079886, + "raw_apical_rms": 0.12062386830639267, + "teaching_rms": 0.0005767004225555261, + "traffic_rms": 0.12062093488199492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.405657788060125e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5847056685689057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01100363751156407, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.0036332080767676233 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011377066373825, + "mean_feedback_forward_cosine": 0.999431684884158, + "mean_feedback_forward_relative_error": 0.03134370776401325, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005560159997639089, + "instruction_rms": 0.000556016000007973, + "raw_apical_rms": 0.1206275819578644, + "teaching_rms": 0.0005560159997639089, + "traffic_rms": 0.12062460473059103 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.179609357559457e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5983546141701658e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010976496711274413, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0036046438322030008 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0701136514544487, + "mean_feedback_forward_cosine": 0.999431679465554, + "mean_feedback_forward_relative_error": 0.03134365017441186, + "min_feedback_forward_cosine": 0.9975558519363403 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006151695874118451, + "instruction_rms": 0.0006151695874295072, + "raw_apical_rms": 0.12063329848227758, + "teaching_rms": 0.0006151695874118451, + "traffic_rms": 0.12062962135253426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.8280051007386646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.594326998629107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011050145131733127, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0038107373747229578 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011356204748154, + "mean_feedback_forward_cosine": 0.9994316859678789, + "mean_feedback_forward_relative_error": 0.03134360311722213, + "min_feedback_forward_cosine": 0.9975559115409851 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005355511792676318, + "instruction_rms": 0.0005355511791900692, + "raw_apical_rms": 0.12062930766844372, + "teaching_rms": 0.0005355511792676318, + "traffic_rms": 0.12062685877656662 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.520752293681653e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.583419186791015e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011023026841280002, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0034027967879734933 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011345028877258, + "mean_feedback_forward_cosine": 0.999431684884158, + "mean_feedback_forward_relative_error": 0.03134355276653712, + "min_feedback_forward_cosine": 0.9975559115409851 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000571202598471784, + "instruction_rms": 0.0005712025982363581, + "raw_apical_rms": 0.12062763988800765, + "teaching_rms": 0.000571202598471784, + "traffic_rms": 0.1206245600619184 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.085708375598188e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587018902721264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011264952180325645, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.0035343935879319907 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011336833238602, + "mean_feedback_forward_cosine": 0.9994316957213661, + "mean_feedback_forward_relative_error": 0.031343505455350334, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005420570398281949, + "instruction_rms": 0.0005420570398467584, + "raw_apical_rms": 0.12062687888434831, + "teaching_rms": 0.0005420570398281949, + "traffic_rms": 0.12062409891134195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 5.9498258764278944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5856916888685357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010932533688843974, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.0033941391864046456 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011322677135468, + "mean_feedback_forward_cosine": 0.999431684884158, + "mean_feedback_forward_relative_error": 0.03134344337846745, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006550410990537408, + "instruction_rms": 0.0006550410987671143, + "raw_apical_rms": 0.12062245567919055, + "teaching_rms": 0.0006550410990537408, + "traffic_rms": 0.12061847060003508 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.08072383556646e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5900002589981593e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01103817963938633, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.004194585617929697 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011310756206512, + "mean_feedback_forward_cosine": 0.9994317000562495, + "mean_feedback_forward_relative_error": 0.03134338492527604, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006129186839902447, + "instruction_rms": 0.0006129186838665845, + "raw_apical_rms": 0.12061725939845196, + "teaching_rms": 0.0006129186839902447, + "traffic_rms": 0.12061381182795533 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.5883788263599854e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5896749282567908e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010985667539256911, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.0038366151857189834 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011300325393677, + "mean_feedback_forward_cosine": 0.9994316924702037, + "mean_feedback_forward_relative_error": 0.031343333330005406, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005434226332477861, + "instruction_rms": 0.0005434226333631546, + "raw_apical_rms": 0.12062086958413583, + "teaching_rms": 0.0005434226332477861, + "traffic_rms": 0.12061814844798308 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.343538234114931e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5843312635637782e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011119251710338651, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0033230470594763756 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011286169290543, + "mean_feedback_forward_cosine": 0.9994316870515997, + "mean_feedback_forward_relative_error": 0.0313432781110433, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006050422177946555, + "instruction_rms": 0.000605042217856636, + "raw_apical_rms": 0.12061878319968548, + "teaching_rms": 0.0006050422177946555, + "traffic_rms": 0.12061509153666344 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.03568900917162e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5884562432828783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011086398436325055, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.0039166533627547325 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011277973651886, + "mean_feedback_forward_cosine": 0.9994316816329956, + "mean_feedback_forward_relative_error": 0.03134323151951486, + "min_feedback_forward_cosine": 0.9975561499595642 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005044759147448961, + "instruction_rms": 0.0005044759146976001, + "raw_apical_rms": 0.12062000297789738, + "teaching_rms": 0.0005044759147448961, + "traffic_rms": 0.12061750861809446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.918615642031e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5854146763837002e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011066784017698842, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.003106705672647804 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011265307664871, + "mean_feedback_forward_cosine": 0.9994316989725286, + "mean_feedback_forward_relative_error": 0.031343181888488204, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005659631940789077, + "instruction_rms": 0.0005659631940980432, + "raw_apical_rms": 0.12061559914493383, + "teaching_rms": 0.0005659631940789077, + "traffic_rms": 0.12061268996074713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.140850127347221e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5862176173483498e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011001812745384202, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.0033360494340211153 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011255621910095, + "mean_feedback_forward_cosine": 0.9994316957213661, + "mean_feedback_forward_relative_error": 0.031343130623413755, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005673543001239207, + "instruction_rms": 0.0005673543001508222, + "raw_apical_rms": 0.1206153567101927, + "teaching_rms": 0.0005673543001239207, + "traffic_rms": 0.12061235260013044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.968820460111601e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.582743846822535e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010983011168720963, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0034580367625225337 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011246681213379, + "mean_feedback_forward_cosine": 0.9994316783818331, + "mean_feedback_forward_relative_error": 0.03134308641945774, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004976459815138841, + "instruction_rms": 0.0004976459816920105, + "raw_apical_rms": 0.12061366400056805, + "teaching_rms": 0.0004976459815138841, + "traffic_rms": 0.12061126926191726 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.905246152655309e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5906991961414833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011195551727493036, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.003201853703260422 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011238485574722, + "mean_feedback_forward_cosine": 0.9994317152283408, + "mean_feedback_forward_relative_error": 0.03134304222396829, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005211968726319222, + "instruction_rms": 0.0005211968725091514, + "raw_apical_rms": 0.12061572611032415, + "teaching_rms": 0.0005211968726319222, + "traffic_rms": 0.12061317084171905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.850057149814347e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5872184464147067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011010572333556977, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.00335953976765275 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011228054761887, + "mean_feedback_forward_cosine": 0.9994317108934576, + "mean_feedback_forward_relative_error": 0.03134299394759265, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005388063977792486, + "instruction_rms": 0.0005388063980582437, + "raw_apical_rms": 0.12061254561114616, + "teaching_rms": 0.0005388063977792486, + "traffic_rms": 0.1206097198136316 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.545095653289536e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5939430196427086e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011011390089305627, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0034279786084406077 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011214643716812, + "mean_feedback_forward_cosine": 0.9994317065585744, + "mean_feedback_forward_relative_error": 0.03134293638339097, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006285226746836842, + "instruction_rms": 0.0006285226751816846, + "raw_apical_rms": 0.12061349649768338, + "teaching_rms": 0.0006285226746836842, + "traffic_rms": 0.12060975869621608 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.66417045383605e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5894972007633913e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011204382048263899, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003654591364916414 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011202722787857, + "mean_feedback_forward_cosine": 0.9994317033074119, + "mean_feedback_forward_relative_error": 0.03134288574484262, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005821155552124259, + "instruction_rms": 0.0005821155549864073, + "raw_apical_rms": 0.12061333577180527, + "teaching_rms": 0.0005821155552124259, + "traffic_rms": 0.1206101166896309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.286761333740287e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5891859237915765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011134620326530023, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003828974887933582 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011193782091141, + "mean_feedback_forward_cosine": 0.9994317054748535, + "mean_feedback_forward_relative_error": 0.03134284430098805, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047242418377295165, + "instruction_rms": 0.00047242418350861443, + "raw_apical_rms": 0.12061052989384974, + "teaching_rms": 0.00047242418377295165, + "traffic_rms": 0.12060850050911034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.890857662256167e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.58954814347586e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01102968120452042, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.0028017387136630716 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011181861162186, + "mean_feedback_forward_cosine": 0.9994316978888078, + "mean_feedback_forward_relative_error": 0.031342793806371366, + "min_feedback_forward_cosine": 0.9975559711456299 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005903892566277716, + "instruction_rms": 0.0005903892564233846, + "raw_apical_rms": 0.12060807956270969, + "teaching_rms": 0.0005903892566277716, + "traffic_rms": 0.1206047511850156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.08429502335639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5874165376027638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011040192633313335, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0038351499533094465 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0701117292046547, + "mean_feedback_forward_cosine": 0.9994317119771784, + "mean_feedback_forward_relative_error": 0.03134274671531536, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005448232452103345, + "instruction_rms": 0.0005448232451741938, + "raw_apical_rms": 0.12060491919545796, + "teaching_rms": 0.0005448232452103345, + "traffic_rms": 0.12060228145556975 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 8.226812298062214e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5861515717165807e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011145206738582146, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.0033393716052360833 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011160999536514, + "mean_feedback_forward_cosine": 0.9994317152283408, + "mean_feedback_forward_relative_error": 0.03134268840605563, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005763343930433735, + "instruction_rms": 0.0005763343930394777, + "raw_apical_rms": 0.12060809345405296, + "teaching_rms": 0.0005763343930433735, + "traffic_rms": 0.12060482781205767 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.131647284883911e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5844696801690102e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011045730086426286, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0036598750675283375 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011149078607559, + "mean_feedback_forward_cosine": 0.9994317098097367, + "mean_feedback_forward_relative_error": 0.03134263750504364, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005744581670031234, + "instruction_rms": 0.0005744581666903161, + "raw_apical_rms": 0.12060959924506831, + "teaching_rms": 0.0005744581670031234, + "traffic_rms": 0.12060653359328767 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 7.117659350797112e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5906953395295424e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011006976178952728, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.0036172117663547396 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011138647794724, + "mean_feedback_forward_cosine": 0.9994317152283408, + "mean_feedback_forward_relative_error": 0.03134259288622574, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005111458291816906, + "instruction_rms": 0.0005111458292987282, + "raw_apical_rms": 0.12060912806148832, + "teaching_rms": 0.0005111458291816906, + "traffic_rms": 0.1206064117856588 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.878908382328518e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.591159473491803e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011075565606534716, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0033742073041945694 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011128216981888, + "mean_feedback_forward_cosine": 0.9994317217306657, + "mean_feedback_forward_relative_error": 0.031342542891136624, + "min_feedback_forward_cosine": 0.9975560307502747 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005398211842836283, + "instruction_rms": 0.0005398211843157291, + "raw_apical_rms": 0.12060749507978928, + "teaching_rms": 0.0005398211842836283, + "traffic_rms": 0.12060478850487091 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.236189875608034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.587612923575866e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01108731780416762, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.0032701864750124513 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011117786169052, + "mean_feedback_forward_cosine": 0.9994317054748535, + "mean_feedback_forward_relative_error": 0.03134249184619297, + "min_feedback_forward_cosine": 0.997556209564209 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000524316796082911, + "instruction_rms": 0.0005243167960490177, + "raw_apical_rms": 0.12060312560575721, + "teaching_rms": 0.000524316796082911, + "traffic_rms": 0.12060031790672346 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 5.935772406928663e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5928149098718884e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01112647086349768, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0033051172525342554 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011107355356216, + "mean_feedback_forward_cosine": 0.9994317130608992, + "mean_feedback_forward_relative_error": 0.03134244018318978, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005683161321633778, + "instruction_rms": 0.0005683161320164979, + "raw_apical_rms": 0.1206061315031053, + "teaching_rms": 0.0005683161321633778, + "traffic_rms": 0.1206031441660092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.613117875531316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.589918210121876e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011103798119106409, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.003474868165999651 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011096179485321, + "mean_feedback_forward_cosine": 0.9994317195632241, + "mean_feedback_forward_relative_error": 0.03134238957004114, + "min_feedback_forward_cosine": 0.9975560903549194 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005417618198583302, + "instruction_rms": 0.0005417618198572267, + "raw_apical_rms": 0.12060498828107377, + "teaching_rms": 0.0005417618198583302, + "traffic_rms": 0.12060221464078606 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260973036289215, + "maximum_absolute_post_projection_soma_slope": 6.244734152005549e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5874611353022428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011158040318562856, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.0033151503371819852 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07011085003614426, + "mean_feedback_forward_cosine": 0.999431696805087, + "mean_feedback_forward_relative_error": 0.03134233936328779, + "min_feedback_forward_cosine": 0.9975561499595642 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005486043037678133, + "instruction_rms": 0.0005486043037548054, + "raw_apical_rms": 0.12060382867848905, + "teaching_rms": 0.0005486043037678133, + "traffic_rms": 0.12060086328985489 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.14260974526405334, + "maximum_absolute_post_projection_soma_slope": 6.818186193413567e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5899783679265844e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.011041197951431369, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.003250983671899885 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9265, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.40451343078613283, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3324800000, + "peak_memory_reserved_bytes": 5553258496, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 6.46296953732417e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 6.46296953732417e-08, + "min_residual_soma_slope": -6.405959140920459e-08, + "mse": 7.551191395303828e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.183399530690697e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.183399530690697e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 13, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.0277559757232666, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.06840205192565918, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 26341.438787698746, + "train_wall_s": 26335.04203605652 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.03324144333600998, + 0.023956647142767906, + 0.016600562259554863, + 0.01499030739068985, + 0.010612402111291885, + 0.012000334449112415, + 0.00762191554531455, + 0.009904524311423302, + 0.005702985916286707, + 0.007403014227747917, + 0.004532030317932367, + 0.006072201766073704, + 0.0036331163719296455, + 0.0051048751920461655, + 0.002998217474669218, + 0.004114323761314154, + 0.0026671418454498053, + 0.003895387751981616, + 0.002370318630710244, + 0.0056512546725571156, + 0.0037113670259714127, + 0.005224796012043953, + 0.0032843619119375944, + 0.004609318450093269, + 0.002866135211661458, + 0.0039026974700391293, + 0.0026347883976995945, + 0.003975309431552887, + 0.0023530018515884876, + 0.0033136510755866766, + 0.0021715739276260138, + 0.0032538925297558308, + 0.001960910391062498, + 0.00299330847337842, + 0.0018442444270476699, + 0.0027256158646196127, + 0.001736381440423429, + 0.004094036761671305, + 0.0028829944785684347, + 0.0039848401211202145, + 0.0026638475246727467, + 0.0036872976925224066, + 0.0024965289048850536, + 0.003642389550805092, + 0.0023068618029356003, + 0.003363950178027153, + 0.0021544655319303274, + 0.0031548282131552696, + 0.0020185247994959354, + 0.002877783728763461, + 0.001951601472683251, + 0.0028882359620183706, + 0.0018873275257647038, + 0.0026884444523602724, + 0.0018550545210018754 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.000000476837158, + 3.999999523162842, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.17709320783615112, + 0.124513179063797, + 0.058669257909059525, + 0.07763158529996872, + 0.028829237446188927, + 0.06371939927339554, + 0.017623847350478172, + 0.052935902029275894, + 0.011839767917990685, + 0.03833375126123428, + 0.008274483494460583, + 0.03276669606566429, + 0.006016725208610296, + 0.027274494990706444, + 0.0045689488761126995, + 0.022393498569726944, + 0.0038804931100457907, + 0.021015234291553497, + 0.0032204589806497097, + 0.029996510595083237, + 0.006462405901402235, + 0.027751736342906952, + 0.005268997512757778, + 0.024546777829527855, + 0.004274462815374136, + 0.02027544006705284, + 0.0036558243446052074, + 0.021232953295111656, + 0.0030691949650645256, + 0.017194336280226707, + 0.0026687579229474068, + 0.017576754093170166, + 0.0022942451760172844, + 0.016171762719750404, + 0.00208999402821064, + 0.014411799609661102, + 0.001927451230585575, + 0.021360954269766808, + 0.004324133042246103, + 0.02077731117606163, + 0.0037467011716216803, + 0.019512439146637917, + 0.0032881060615181923, + 0.01938728243112564, + 0.002971186302602291, + 0.01791982166469097, + 0.0026235159020870924, + 0.01685306988656521, + 0.0023501201067119837, + 0.015273039229214191, + 0.0022179309744387865, + 0.015392888337373734, + 0.0020608259364962578, + 0.014193649403750896, + 0.0019489242695271969 + ], + "unscaled_traffic_rms": [ + 0.7508236765861511, + 0.7696099877357483, + 1.131806492805481, + 0.7723819017410278, + 1.4724498987197876, + 0.7533237934112549, + 1.7299095392227173, + 0.7484164237976074, + 1.9267221689224243, + 0.7724800109863281, + 2.1908462047576904, + 0.741265058517456, + 2.415344715118408, + 0.748666524887085, + 2.624864101409912, + 0.7349139451980591, + 2.749281406402588, + 0.7414407730102539, + 2.944075584411621, + 0.7535882592201233, + 2.2972044944763184, + 0.7530766129493713, + 2.4933485984802246, + 0.7511076927185059, + 2.682101011276245, + 0.7699359059333801, + 2.8828392028808594, + 0.7488943338394165, + 3.0666046142578125, + 0.7708703279495239, + 3.2548084259033203, + 0.7404990792274475, + 3.4188332557678223, + 0.7403790354728699, + 3.5296645164489746, + 0.7564956545829773, + 3.6034767627716064, + 0.7666392922401428, + 2.6668879985809326, + 0.7671522498130798, + 2.8439390659332275, + 0.7558865547180176, + 3.0370419025421143, + 0.7515007853507996, + 3.1056439876556396, + 0.750889241695404, + 3.2848522663116455, + 0.7487841844558716, + 3.4356112480163574, + 0.7536898851394653, + 3.519679546356201, + 0.7505377531051636, + 3.6632449626922607, + 0.7576471567153931, + 3.807340383529663 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250441282772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 127796290519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 5010596513857536 + } +} diff --git a/results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json new file mode 100644 index 0000000..84fc0ac --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json @@ -0,0 +1,7293 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 0, + "eval_split": "test", + "learn_P": 1, + "loader_seed": 14, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5014, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 10000000, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9994279046853384, + "feedback_forward_cosine": [ + 0.9997988343238831, + 0.9997749328613281, + 0.9991828799247742, + 0.999111533164978, + 0.999463677406311, + 0.9993107914924622, + 0.9992161989212036, + 0.9990545511245728, + 0.998909592628479, + 0.9987071752548218, + 0.999567985534668, + 0.9993953704833984, + 0.9997379779815674, + 0.9995555877685547, + 0.9996947050094604, + 0.999505341053009, + 0.9994285106658936, + 0.9991961717605591, + 0.9993981719017029, + 0.9995056390762329, + 0.9992136359214783, + 0.9991928339004517, + 0.9990469217300415, + 0.9990625381469727, + 0.9992903470993042, + 0.9992948770523071, + 0.9995030164718628, + 0.9994757771492004, + 0.9996125102043152, + 0.9995819330215454, + 0.999678909778595, + 0.9996166825294495, + 0.9997760057449341, + 0.9996891021728516, + 0.9997944831848145, + 0.9996936321258545, + 0.9997571110725403, + 0.9998501539230347, + 0.9998060464859009, + 0.9997848868370056, + 0.999779462814331, + 0.9997221827507019, + 0.9997811913490295, + 0.9996858835220337, + 0.9997216463088989, + 0.9995635747909546, + 0.9996534585952759, + 0.999471127986908, + 0.9996280670166016, + 0.9994121193885803, + 0.9996070861816406, + 0.9993250370025635, + 0.9995989203453064, + 0.9991874098777771, + 0.9999798536300659 + ], + "feedback_forward_norm_ratio": [ + 1.000149130821228, + 1.0009535551071167, + 0.9984815120697021, + 1.0016766786575317, + 0.9995856881141663, + 1.0017879009246826, + 0.9959774613380432, + 1.0028762817382812, + 0.9986860156059265, + 1.0021672248840332, + 0.9988604784011841, + 1.0024598836898804, + 0.9981681108474731, + 1.0017732381820679, + 0.9997989535331726, + 1.0013729333877563, + 0.9980446100234985, + 1.0017657279968262, + 0.9982770085334778, + 1.0011591911315918, + 0.9984641671180725, + 1.0009673833847046, + 0.9978606700897217, + 1.0017354488372803, + 0.9988399147987366, + 1.0011192560195923, + 0.9991135597229004, + 1.0013725757598877, + 0.9988991022109985, + 1.0011109113693237, + 0.999144971370697, + 1.0011439323425293, + 0.9994403123855591, + 1.000425934791565, + 0.9993515014648438, + 1.0012404918670654, + 0.9997034072875977, + 1.0002387762069702, + 0.9998688101768494, + 1.000625491142273, + 0.9995481371879578, + 1.0008283853530884, + 0.9997639060020447, + 1.0010287761688232, + 0.9995730519294739, + 1.00138258934021, + 0.9991338849067688, + 1.001706838607788, + 0.9990312457084656, + 1.0018060207366943, + 0.9994123578071594, + 1.0015783309936523, + 0.9993245005607605, + 1.0016975402832031, + 1.0021151304244995 + ], + "feedback_forward_relative_error": [ + 0.020060308277606964, + 0.021246453747153282, + 0.04042421653866768, + 0.04222222790122032, + 0.03274479880928993, + 0.037205327302217484, + 0.03971664980053902, + 0.04364164546132088, + 0.04668582230806351, + 0.05095241591334343, + 0.02939867228269577, + 0.034902483224868774, + 0.022942183539271355, + 0.029894202947616577, + 0.02470928616821766, + 0.03150338679552078, + 0.0338282510638237, + 0.04016990214586258, + 0.03470529988408089, + 0.031483158469200134, + 0.039652612060308456, + 0.04020845890045166, + 0.04366402328014374, + 0.043372295796871185, + 0.0376720055937767, + 0.037591613829135895, + 0.03152763843536377, + 0.032432928681373596, + 0.027845030650496483, + 0.028955554589629173, + 0.02534536086022854, + 0.027727821841835976, + 0.02116387151181698, + 0.024943169206380844, + 0.020272381603717804, + 0.024798842146992683, + 0.022038796916604042, + 0.017315328121185303, + 0.019690556451678276, + 0.02075815573334694, + 0.02099420502781868, + 0.023600945249199867, + 0.020909609273076057, + 0.025102714076638222, + 0.023593297228217125, + 0.02959650196135044, + 0.026328735053539276, + 0.032597433775663376, + 0.027278052642941475, + 0.03437091037631035, + 0.028031418099999428, + 0.03680349513888359, + 0.028323985636234283, + 0.04038499668240547, + 0.0066958083771169186 + ], + "innovation_negative_gradient_cosine": [ + 0.9995262622833252, + 0.9994668960571289, + 0.9995425343513489, + 0.9991871118545532, + 0.9995458126068115, + 0.9992923736572266, + 0.9995535016059875, + 0.9990665316581726, + 0.9995579719543457, + 0.9990013837814331, + 0.9995593428611755, + 0.9994300603866577, + 0.9995750188827515, + 0.9994897246360779, + 0.9995806217193604, + 0.9994637370109558, + 0.9995887279510498, + 0.9992746710777283, + 0.9995965957641602, + 0.9993570446968079, + 0.9996023774147034, + 0.9992549419403076, + 0.9996078014373779, + 0.9991986751556396, + 0.9996156096458435, + 0.9992837905883789, + 0.9996231198310852, + 0.9993802309036255, + 0.9996287822723389, + 0.9994536638259888, + 0.9996305704116821, + 0.9995105266571045, + 0.9996395111083984, + 0.9995378255844116, + 0.999650239944458, + 0.9995735287666321, + 0.9996629357337952, + 0.9996985197067261, + 0.9996863603591919, + 0.9996581673622131, + 0.9997047185897827, + 0.9996294379234314, + 0.9997169971466064, + 0.9996535778045654, + 0.9997180104255676, + 0.9996700286865234, + 0.9997422099113464, + 0.9996808767318726, + 0.9997777342796326, + 0.9996774196624756, + 0.9998447299003601, + 0.9998321533203125, + 0.9999211430549622, + 0.9999521970748901, + 0.8918465375900269 + ], + "instruction_negative_gradient_cosine": [ + 0.99952632188797, + 0.9994668960571289, + 0.9995425343513489, + 0.9991871118545532, + 0.9995458126068115, + 0.9992923736572266, + 0.9995535612106323, + 0.9990665316581726, + 0.9995579719543457, + 0.9990013837814331, + 0.9995593428611755, + 0.9994300007820129, + 0.9995750188827515, + 0.9994897246360779, + 0.9995806217193604, + 0.999463677406311, + 0.9995887279510498, + 0.999274730682373, + 0.9995965957641602, + 0.9993569850921631, + 0.9996023774147034, + 0.9992550015449524, + 0.9996078014373779, + 0.9991986751556396, + 0.9996156692504883, + 0.9992837309837341, + 0.9996230602264404, + 0.9993802905082703, + 0.9996287822723389, + 0.9994536638259888, + 0.9996305704116821, + 0.9995105266571045, + 0.9996395111083984, + 0.9995378255844116, + 0.999650239944458, + 0.9995734691619873, + 0.9996629953384399, + 0.9996985197067261, + 0.9996863603591919, + 0.9996581673622131, + 0.9997045993804932, + 0.9996294975280762, + 0.9997169375419617, + 0.9996535778045654, + 0.9997179508209229, + 0.9996700286865234, + 0.9997422099113464, + 0.9996808767318726, + 0.9997777342796326, + 0.9996774196624756, + 0.9998446702957153, + 0.9998321533203125, + 0.9999211430549622, + 0.9999521970748901, + 0.9127292037010193 + ], + "matched_negative_gradient_cosine": [ + -0.008187413215637207, + -0.015598754398524761, + -0.011859042569994926, + -0.00952020101249218, + -0.009076099842786789, + 0.025975488126277924, + -0.006392987444996834, + -0.025431979447603226, + -0.0024859346449375153, + -0.024633590131998062, + -0.003860069904476404, + 0.027092572301626205, + -0.01832091063261032, + -0.037103861570358276, + -0.011960268020629883, + -0.020264824852347374, + -0.013413629494607449, + 0.019100390374660492, + -0.011545455083251, + 0.04292016476392746, + -0.03241690993309021, + 0.0434187576174736, + -0.04402662441134453, + 0.05480309948325157, + -0.046772345900535583, + 0.031519565731287, + -0.045361995697021484, + 0.036612339317798615, + -0.049954161047935486, + 0.05757954716682434, + -0.046010348945856094, + 0.03367879241704941, + -0.03750961273908615, + 0.024862172082066536, + -0.028193483129143715, + 0.04454458877444267, + -0.030192747712135315, + -0.04034389555454254, + -0.09494373202323914, + 0.038277141749858856, + -0.10235662758350372, + 0.04242897778749466, + -0.10648664832115173, + 0.03178362920880318, + -0.07564844191074371, + 0.07797741889953613, + -0.051776088774204254, + 0.15095344185829163, + -0.08517353981733322, + 0.0781029760837555, + -0.06110972911119461, + 0.004467208869755268, + -0.021361134946346283, + -0.08613333106040955, + 0.09746018052101135 + ], + "max_norm_match_direction_error": 0.23414266109466553, + "max_norm_match_relative_error": 1.710782839836611e-07, + "mean_feedback_forward_cosine": 0.9994973464445635, + "mean_feedback_forward_relative_error": 0.03054591358893297, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.3654601275920868, + "max_absolute_post_projection_soma_slope": 4.1842426412586065e-08, + "max_absolute_pre_projection_soma_slope": 0.3654601275920868, + "max_positive_post_projection_soma_slope": 4.1842426412586065e-08, + "min_post_projection_soma_slope": -2.896273976205066e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.3365609282722773e-08, + "pre_projection_traffic_rms_ratio": 0.011779218723666182 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.011779218723648404, + "raw_negative_gradient_cosine": [ + -0.008187413215637207, + -0.015598755329847336, + -0.011859041638672352, + -0.00952020101249218, + -0.009076099842786789, + 0.025975488126277924, + -0.0063929869793355465, + -0.025431979447603226, + -0.0024859346449375153, + -0.024633590131998062, + -0.0038600703701376915, + 0.027092570438981056, + -0.01832091063261032, + -0.037103861570358276, + -0.011960268020629883, + -0.020264824852347374, + -0.013413628563284874, + 0.019100388512015343, + -0.011545455083251, + 0.04292016476392746, + -0.03241691365838051, + 0.0434187576174736, + -0.04402662441134453, + 0.05480309948325157, + -0.046772342175245285, + 0.031519565731287, + -0.04536199942231178, + 0.03661234304308891, + -0.049954161047935486, + 0.057579539716243744, + -0.046010348945856094, + 0.03367879241704941, + -0.03750961273908615, + 0.024862170219421387, + -0.028193486854434013, + 0.04454458877444267, + -0.030192747712135315, + -0.04034389927983284, + -0.09494373202323914, + 0.038277141749858856, + -0.10235662758350372, + 0.04242897778749466, + -0.10648664832115173, + 0.03178362548351288, + -0.07564844191074371, + 0.07797741889953613, + -0.051776085048913956, + 0.15095345675945282, + -0.08517353981733322, + 0.0781029686331749, + -0.06110972911119461, + 0.004467210732400417, + -0.021361136808991432, + -0.08613333106040955, + 0.09750264883041382 + ], + "teaching_negative_gradient_cosine": [ + 0.9995262622833252, + 0.9994668960571289, + 0.9995425343513489, + 0.9991871118545532, + 0.9995458126068115, + 0.9992923736572266, + 0.9995535016059875, + 0.9990665316581726, + 0.9995579719543457, + 0.9990013837814331, + 0.9995593428611755, + 0.9994300603866577, + 0.9995750188827515, + 0.9994897246360779, + 0.9995806217193604, + 0.9994637370109558, + 0.9995887279510498, + 0.9992746710777283, + 0.9995965957641602, + 0.9993570446968079, + 0.9996023774147034, + 0.9992549419403076, + 0.9996078014373779, + 0.9991986751556396, + 0.9996156096458435, + 0.9992837905883789, + 0.9996231198310852, + 0.9993802309036255, + 0.9996287822723389, + 0.9994536638259888, + 0.9996305704116821, + 0.9995105266571045, + 0.9996395111083984, + 0.9995378255844116, + 0.999650239944458, + 0.9995735287666321, + 0.9996629357337952, + 0.9996985197067261, + 0.9996863603591919, + 0.9996581673622131, + 0.9997047185897827, + 0.9996294379234314, + 0.9997169971466064, + 0.9996535778045654, + 0.9997180104255676, + 0.9996700286865234, + 0.9997422099113464, + 0.9996808767318726, + 0.9997777342796326, + 0.9996774196624756, + 0.9998447299003601, + 0.9998321533203125, + 0.9999211430549622, + 0.9999521970748901, + 0.8918465375900269 + ], + "traffic_instruction_rms_ratio": 289.5474469652048, + "used_negative_gradient_cosine": [ + 0.9995262622833252, + 0.9994668960571289, + 0.9995425343513489, + 0.9991871118545532, + 0.9995458126068115, + 0.9992923736572266, + 0.9995535016059875, + 0.9990665316581726, + 0.9995579719543457, + 0.9990013837814331, + 0.9995593428611755, + 0.9994300603866577, + 0.9995750188827515, + 0.9994897246360779, + 0.9995806217193604, + 0.9994637370109558, + 0.9995887279510498, + 0.9992746710777283, + 0.9995965957641602, + 0.9993570446968079, + 0.9996023774147034, + 0.9992549419403076, + 0.9996078014373779, + 0.9991986751556396, + 0.9996156096458435, + 0.9992837905883789, + 0.9996231198310852, + 0.9993802309036255, + 0.9996287822723389, + 0.9994536638259888, + 0.9996305704116821, + 0.9995105266571045, + 0.9996395111083984, + 0.9995378255844116, + 0.999650239944458, + 0.9995735287666321, + 0.9996629357337952, + 0.9996985197067261, + 0.9996863603591919, + 0.9996581673622131, + 0.9997047185897827, + 0.9996294379234314, + 0.9997169971466064, + 0.9996535778045654, + 0.9997180104255676, + 0.9996700286865234, + 0.9997422099113464, + 0.9996808767318726, + 0.9997777342796326, + 0.9996774196624756, + 0.9998447299003601, + 0.9998321533203125, + 0.9999211430549622, + 0.9999521970748901, + 0.8918465375900269 + ], + "wall_s": 0.3509204387664795 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.337538719177246, + "mean_feedback_forward_cosine": 0.3470919665965167, + "mean_feedback_forward_relative_error": 1.1312912729653446, + "min_feedback_forward_cosine": 0.11561739444732666 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0015407984580144906, + "instruction_rms": 0.0015407984577723273, + "raw_apical_rms": 0.06508568969197852, + "teaching_rms": 0.0015407984580144906, + "traffic_rms": 0.06503976337833124 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546027660369873, + "maximum_absolute_post_projection_soma_slope": 1.4074466037072852e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.031357188939468e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015969962131895204, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.318244298744202 + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3209714889526367, + "mean_feedback_forward_cosine": 0.36345010128888217, + "mean_feedback_forward_relative_error": 1.1174056324091826, + "min_feedback_forward_cosine": 0.1323525458574295 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0015949982496293853, + "instruction_rms": 0.0015949982499202469, + "raw_apical_rms": 0.07291536069493962, + "teaching_rms": 0.0015949982496293853, + "traffic_rms": 0.07289750933437934 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.2117673489156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7559230056188866e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014829932795428267, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 1.6927503633880616 + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3107725381851196, + "mean_feedback_forward_cosine": 0.38829824165864424, + "mean_feedback_forward_relative_error": 1.0960471938956868, + "min_feedback_forward_cosine": 0.14350982010364532 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0019947406849959606, + "instruction_rms": 0.0019947406856681052, + "raw_apical_rms": 0.08211166025269341, + "teaching_rms": 0.0019947406849959606, + "traffic_rms": 0.08208711997049112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.589142990762412e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6735927927764367e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.014280128604741968, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 1.4528260779571534 + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2989331483840942, + "mean_feedback_forward_cosine": 0.41964017044414176, + "mean_feedback_forward_relative_error": 1.0679928118532354, + "min_feedback_forward_cosine": 0.15901969373226166 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0023721994444680056, + "instruction_rms": 0.0023721994443365716, + "raw_apical_rms": 0.09075613528587424, + "teaching_rms": 0.0023721994444680056, + "traffic_rms": 0.0907247836261394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.099108666532629e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6259860321962358e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01644804706433687, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 1.2662530714035034 + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2839618921279907, + "mean_feedback_forward_cosine": 0.4521103149110621, + "mean_feedback_forward_relative_error": 1.0377678621898998, + "min_feedback_forward_cosine": 0.17882689833641052 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00263376982902271, + "instruction_rms": 0.0026337698286168086, + "raw_apical_rms": 0.09789696295851179, + "teaching_rms": 0.00263376982902271, + "traffic_rms": 0.09786118646541846 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.615807528007281e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6011019227903277e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01908546054893131, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 1.10482550365448 + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2659108638763428, + "mean_feedback_forward_cosine": 0.4828754099932584, + "mean_feedback_forward_relative_error": 1.0081115467981858, + "min_feedback_forward_cosine": 0.20184218883514404 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027890276042216436, + "instruction_rms": 0.0027890276045768104, + "raw_apical_rms": 0.10561079437773366, + "teaching_rms": 0.0027890276042216436, + "traffic_rms": 0.10557362962319122 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.142576118392753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5842031667546104e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02031041063392772, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 0.970712696723938 + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2453035116195679, + "mean_feedback_forward_cosine": 0.5108968458392403, + "mean_feedback_forward_relative_error": 0.9799336552619934, + "min_feedback_forward_cosine": 0.22743552923202515 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027672997998718263, + "instruction_rms": 0.002767299800172459, + "raw_apical_rms": 0.11128087703751355, + "teaching_rms": 0.0027672997998718263, + "traffic_rms": 0.1112461774457811 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.724236977286637e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5703843215636337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.020073828557752606, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 0.8659917796325683 + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2211562395095825, + "mean_feedback_forward_cosine": 0.537326720086011, + "mean_feedback_forward_relative_error": 0.9524023408239538, + "min_feedback_forward_cosine": 0.2589121460914612 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028219036240894934, + "instruction_rms": 0.0028219036245949397, + "raw_apical_rms": 0.11573784220686462, + "teaching_rms": 0.0028219036240894934, + "traffic_rms": 0.11570313540104829 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.4154767313812044e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5620650117810692e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.02082237532851249, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 0.770338599395752 + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1937166452407837, + "mean_feedback_forward_cosine": 0.5625516674735329, + "mean_feedback_forward_relative_error": 0.9254404420202429, + "min_feedback_forward_cosine": 0.29125499725341797 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002760851634095237, + "instruction_rms": 0.00276085163384671, + "raw_apical_rms": 0.1194780064929881, + "teaching_rms": 0.002760851634095237, + "traffic_rms": 0.1194457704314566 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.2057840592660796e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5632488265037338e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01837243069286356, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 0.7051988785171509 + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.165170431137085, + "mean_feedback_forward_cosine": 0.5863702877001329, + "mean_feedback_forward_relative_error": 0.8992986023426056, + "min_feedback_forward_cosine": 0.32406896352767944 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027336345599632862, + "instruction_rms": 0.0027336345606156338, + "raw_apical_rms": 0.12338395978319029, + "teaching_rms": 0.0027336345599632862, + "traffic_rms": 0.12335339617673681 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1792309351221775e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5437112549538877e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017950120557470466, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 0.653767914428711 + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1369038820266724, + "mean_feedback_forward_cosine": 0.6088971907442267, + "mean_feedback_forward_relative_error": 0.8738858862356707, + "min_feedback_forward_cosine": 0.35734426975250244 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002730209954120157, + "instruction_rms": 0.002730209953863333, + "raw_apical_rms": 0.126435420815262, + "teaching_rms": 0.002730209954120157, + "traffic_rms": 0.12640562411197157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.2467593535347987e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.540149448355781e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017401858975393276, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 0.5981809624862671 + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.105292558670044, + "mean_feedback_forward_cosine": 0.6304948416623202, + "mean_feedback_forward_relative_error": 0.8489244759082795, + "min_feedback_forward_cosine": 0.39154842495918274 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027087560493800782, + "instruction_rms": 0.0027087560492602414, + "raw_apical_rms": 0.12933379945287843, + "teaching_rms": 0.0027087560493800782, + "traffic_rms": 0.12930511500706984 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.1071247030258746e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5361549125476262e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01876124656157677, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 0.5689019970703125 + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0798101425170898, + "mean_feedback_forward_cosine": 0.6512302041053772, + "mean_feedback_forward_relative_error": 0.8240893071348017, + "min_feedback_forward_cosine": 0.4279969334602356 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002707850239468869, + "instruction_rms": 0.0027078502396314704, + "raw_apical_rms": 0.13171015243354006, + "teaching_rms": 0.002707850239468869, + "traffic_rms": 0.13168195856328452 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.418575075414992e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5353506734954773e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016847394984415145, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 0.5342622586250305 + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0569343566894531, + "mean_feedback_forward_cosine": 0.6707149229266427, + "mean_feedback_forward_relative_error": 0.8002192318439484, + "min_feedback_forward_cosine": 0.4579431712627411 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002682962505570068, + "instruction_rms": 0.0026829625059205525, + "raw_apical_rms": 0.13368951031841267, + "teaching_rms": 0.002682962505570068, + "traffic_rms": 0.13366227618650067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.2321903852807736e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.529790427888221e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017538669326089593, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 0.5035887927818299 + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0342109203338623, + "mean_feedback_forward_cosine": 0.6900240085341713, + "mean_feedback_forward_relative_error": 0.7759520229968158, + "min_feedback_forward_cosine": 0.4850977063179016 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00273497832834487, + "instruction_rms": 0.002734978327654097, + "raw_apical_rms": 0.13419910675104824, + "teaching_rms": 0.00273497832834487, + "traffic_rms": 0.13417089811035315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0091090274499948e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5332252161636894e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018612637735523786, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 0.47962559954643247 + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.009242296218872, + "mean_feedback_forward_cosine": 0.7081426035274159, + "mean_feedback_forward_relative_error": 0.752419283173301, + "min_feedback_forward_cosine": 0.5108439326286316 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027413994875524796, + "instruction_rms": 0.0027413994879295954, + "raw_apical_rms": 0.13569418582659817, + "teaching_rms": 0.0027413994875524796, + "traffic_rms": 0.13566609703480437 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0459741872637096e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359596556000106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018380328202013685, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 0.4577398926353455 + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.983150839805603, + "mean_feedback_forward_cosine": 0.725501561164856, + "mean_feedback_forward_relative_error": 0.7291488270867955, + "min_feedback_forward_cosine": 0.5363930463790894 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00271051409466183, + "instruction_rms": 0.002710514094032697, + "raw_apical_rms": 0.13771186833594265, + "teaching_rms": 0.00271051409466183, + "traffic_rms": 0.13768488248457417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.292827485296584e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5273206720856484e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017730250939874244, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 0.4444267007637024 + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.956661581993103, + "mean_feedback_forward_cosine": 0.7419148455966603, + "mean_feedback_forward_relative_error": 0.7065041409297423, + "min_feedback_forward_cosine": 0.5622117519378662 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002710089914123729, + "instruction_rms": 0.002710089913905854, + "raw_apical_rms": 0.13928479352526538, + "teaching_rms": 0.002710089914123729, + "traffic_rms": 0.13925801269541752 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.674059953113101e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5233221870103848e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016508849638922386, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 0.415285579457283 + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9284818172454834, + "mean_feedback_forward_cosine": 0.7578128337860107, + "mean_feedback_forward_relative_error": 0.6839349229227413, + "min_feedback_forward_cosine": 0.5880800485610962 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027264442050400793, + "instruction_rms": 0.002726444204730492, + "raw_apical_rms": 0.1404623510075239, + "teaching_rms": 0.0027264442050400793, + "traffic_rms": 0.14043547964933126 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.795699185133344e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5209482754294717e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017349221786818265, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 0.40599066507339476 + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.901059627532959, + "mean_feedback_forward_cosine": 0.7728046016259627, + "mean_feedback_forward_relative_error": 0.6619364952499216, + "min_feedback_forward_cosine": 0.6123119592666626 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027195105119602156, + "instruction_rms": 0.0027195105115537916, + "raw_apical_rms": 0.1415707072417636, + "teaching_rms": 0.0027195105119602156, + "traffic_rms": 0.1415441535963998 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.224054053902364e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.521941957327478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017178475453871647, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 0.38727013387680054 + }, + { + "epoch": 21, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8736644387245178, + "mean_feedback_forward_cosine": 0.786713340065696, + "mean_feedback_forward_relative_error": 0.6409310308369723, + "min_feedback_forward_cosine": 0.6360628008842468 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026669520435226653, + "instruction_rms": 0.002666952042936354, + "raw_apical_rms": 0.14283187272273373, + "teaching_rms": 0.0026669520435226653, + "traffic_rms": 0.14280654098667978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.192387295797744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5306289243248074e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017102228117365215, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 0.378113029088974 + }, + { + "epoch": 22, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8466385006904602, + "mean_feedback_forward_cosine": 0.7997828711162914, + "mean_feedback_forward_relative_error": 0.6204818468202244, + "min_feedback_forward_cosine": 0.6583391427993774 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027074591583232267, + "instruction_rms": 0.0027074591581023343, + "raw_apical_rms": 0.14392373191092994, + "teaching_rms": 0.0027074591583232267, + "traffic_rms": 0.14389790030024566 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.194648953325668e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.524707498051767e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01699580759902096, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 0.3607297075462341 + }, + { + "epoch": 23, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8186596632003784, + "mean_feedback_forward_cosine": 0.8127056457779624, + "mean_feedback_forward_relative_error": 0.5997356943108819, + "min_feedback_forward_cosine": 0.6804730892181396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027288635420053874, + "instruction_rms": 0.002728863542013064, + "raw_apical_rms": 0.14507399648224498, + "teaching_rms": 0.0027288635420053874, + "traffic_rms": 0.1450479036818884 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.896668728970326e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5242402875867515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017912699478155884, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 0.3533772416782379 + }, + { + "epoch": 24, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7914324998855591, + "mean_feedback_forward_cosine": 0.8245934735644947, + "mean_feedback_forward_relative_error": 0.5799183818427, + "min_feedback_forward_cosine": 0.6965500712394714 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026990716211328343, + "instruction_rms": 0.0026990716211888212, + "raw_apical_rms": 0.14607854508440238, + "teaching_rms": 0.0026990716211328343, + "traffic_rms": 0.14605313903378994 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0374138526003662e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5187876466508386e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01865115340124731, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 0.3402723709487915 + }, + { + "epoch": 25, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7660810947418213, + "mean_feedback_forward_cosine": 0.835796798359264, + "mean_feedback_forward_relative_error": 0.5606765692884271, + "min_feedback_forward_cosine": 0.7125788331031799 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026833759802138706, + "instruction_rms": 0.002683375980144104, + "raw_apical_rms": 0.14707975738516807, + "teaching_rms": 0.0026833759802138706, + "traffic_rms": 0.14705482245630266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.013774615596049e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5208105099546095e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018353248465093543, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 0.32941348722457886 + }, + { + "epoch": 26, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7429596185684204, + "mean_feedback_forward_cosine": 0.846600924838673, + "mean_feedback_forward_relative_error": 0.541546061093157, + "min_feedback_forward_cosine": 0.7291401624679565 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026689660456717538, + "instruction_rms": 0.0026689660460679073, + "raw_apical_rms": 0.14860996429113496, + "teaching_rms": 0.0026689660456717538, + "traffic_rms": 0.14858557100350725 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.758792802154858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5248829977400117e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018314011899773625, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 0.32121589488983154 + }, + { + "epoch": 27, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7203114032745361, + "mean_feedback_forward_cosine": 0.8566695679317821, + "mean_feedback_forward_relative_error": 0.5230962132865732, + "min_feedback_forward_cosine": 0.7445104718208313 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002654322819903756, + "instruction_rms": 0.002654322820730754, + "raw_apical_rms": 0.14890816287616587, + "teaching_rms": 0.002654322819903756, + "traffic_rms": 0.14888409682132533 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0429616992269075e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5236879316142876e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01710572997343234, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 0.31315590534687043 + }, + { + "epoch": 28, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6972556114196777, + "mean_feedback_forward_cosine": 0.8662110274488276, + "mean_feedback_forward_relative_error": 0.5050151432102377, + "min_feedback_forward_cosine": 0.7603663206100464 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026975720280145543, + "instruction_rms": 0.0026975720278505652, + "raw_apical_rms": 0.1496516499756971, + "teaching_rms": 0.0026975720280145543, + "traffic_rms": 0.1496268415406874 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546027660369873, + "maximum_absolute_post_projection_soma_slope": 1.4248362845137308e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5214503290290852e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017181042304619414, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 0.30130046340942385 + }, + { + "epoch": 29, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6768556833267212, + "mean_feedback_forward_cosine": 0.8753467939116738, + "mean_feedback_forward_relative_error": 0.4870706994425167, + "min_feedback_forward_cosine": 0.7736822962760925 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002705575659777348, + "instruction_rms": 0.0027055756596773917, + "raw_apical_rms": 0.15076554569763478, + "teaching_rms": 0.002705575659777348, + "traffic_rms": 0.15074075318198077 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.68105169363298e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5164705112196084e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016756180483932518, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 0.3007733417701721 + }, + { + "epoch": 30, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6617474555969238, + "mean_feedback_forward_cosine": 0.883412655917081, + "mean_feedback_forward_relative_error": 0.4707008208740841, + "min_feedback_forward_cosine": 0.7860751152038574 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026362354965129, + "instruction_rms": 0.0026362354969701914, + "raw_apical_rms": 0.15129607055402097, + "teaching_rms": 0.0026362354965129, + "traffic_rms": 0.15127260812689627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.936544304560812e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5113833384889583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017376240341383145, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 0.2852468093204498 + }, + { + "epoch": 31, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6453670263290405, + "mean_feedback_forward_cosine": 0.8911966042085128, + "mean_feedback_forward_relative_error": 0.4544305476275357, + "min_feedback_forward_cosine": 0.7984600067138672 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00264191337782919, + "instruction_rms": 0.0026419133772344686, + "raw_apical_rms": 0.15220964553605817, + "teaching_rms": 0.00264191337782919, + "traffic_rms": 0.15218613601517356 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.299438374910096e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5145088374094374e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016650105739784586, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 0.27410987197875974 + }, + { + "epoch": 32, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.628040075302124, + "mean_feedback_forward_cosine": 0.8990429672327909, + "mean_feedback_forward_relative_error": 0.43744446716525337, + "min_feedback_forward_cosine": 0.8114415407180786 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002722089965010536, + "instruction_rms": 0.002722089964603603, + "raw_apical_rms": 0.15303129747643762, + "teaching_rms": 0.002722089965010536, + "traffic_rms": 0.15300666015869763 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.945220031364443e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5106566723129036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01735642410266102, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 0.280466178150177 + }, + { + "epoch": 33, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6103391647338867, + "mean_feedback_forward_cosine": 0.9060498270121488, + "mean_feedback_forward_relative_error": 0.42166498777541245, + "min_feedback_forward_cosine": 0.8219460248947144 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002676962300927409, + "instruction_rms": 0.0026769623010211574, + "raw_apical_rms": 0.15388004382506998, + "teaching_rms": 0.002676962300927409, + "traffic_rms": 0.15385625696338195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0543753603542427e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.506974194530283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01662988068859521, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 0.27099272607803343 + }, + { + "epoch": 34, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5925960540771484, + "mean_feedback_forward_cosine": 0.9125872232697226, + "mean_feedback_forward_relative_error": 0.4064791863614863, + "min_feedback_forward_cosine": 0.8332748413085938 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026600693078532617, + "instruction_rms": 0.002660069307873809, + "raw_apical_rms": 0.15526800680664973, + "teaching_rms": 0.0026600693078532617, + "traffic_rms": 0.1552446657813217 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.59805532829705e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.509750249220072e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016071889988459154, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 0.26346910413742064 + }, + { + "epoch": 35, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5764073133468628, + "mean_feedback_forward_cosine": 0.9186815131794323, + "mean_feedback_forward_relative_error": 0.3918265531008894, + "min_feedback_forward_cosine": 0.8417103290557861 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026477325447377673, + "instruction_rms": 0.0026477325443021557, + "raw_apical_rms": 0.1555367961661697, + "teaching_rms": 0.0026477325447377673, + "traffic_rms": 0.15551377053738955 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.535033029806073e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5098573142160327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01620822832611706, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 0.25893822303295133 + }, + { + "epoch": 36, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5594467520713806, + "mean_feedback_forward_cosine": 0.9244397434321316, + "mean_feedback_forward_relative_error": 0.37742655710740525, + "min_feedback_forward_cosine": 0.8506953716278076 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026709120489579016, + "instruction_rms": 0.00267091204900426, + "raw_apical_rms": 0.1568014567908609, + "teaching_rms": 0.0026709120489579016, + "traffic_rms": 0.15677815085454494 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.433845349349212e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5020451980514063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017126357681836114, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 0.25122486134529115 + }, + { + "epoch": 37, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5440977811813354, + "mean_feedback_forward_cosine": 0.9296147519891912, + "mean_feedback_forward_relative_error": 0.36402947509830647, + "min_feedback_forward_cosine": 0.8586518168449402 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026347103646921785, + "instruction_rms": 0.0026347103645644326, + "raw_apical_rms": 0.15757572662308567, + "teaching_rms": 0.0026347103646921785, + "traffic_rms": 0.15755311960772492 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.02934473856476e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.503068806120538e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017803130381265415, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 0.24129646183013917 + }, + { + "epoch": 38, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5284150838851929, + "mean_feedback_forward_cosine": 0.9347350824962962, + "mean_feedback_forward_relative_error": 0.3503709607503631, + "min_feedback_forward_cosine": 0.8666711449623108 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026383876423883633, + "instruction_rms": 0.0026383876431558145, + "raw_apical_rms": 0.15819227002068129, + "teaching_rms": 0.0026383876423883633, + "traffic_rms": 0.15816977197625898 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1452809189904656e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5043882009674355e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017138982974661807, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 0.2524702130126953 + }, + { + "epoch": 39, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.513145923614502, + "mean_feedback_forward_cosine": 0.9392989017746666, + "mean_feedback_forward_relative_error": 0.3377048610286279, + "min_feedback_forward_cosine": 0.8743120431900024 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002619487537411607, + "instruction_rms": 0.0026194875372465747, + "raw_apical_rms": 0.15889544792268626, + "teaching_rms": 0.002619487537411607, + "traffic_rms": 0.15887334345948018 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.192317662609639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5095170440596577e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016600176170958665, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 0.23605706344604493 + }, + { + "epoch": 40, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4964372217655182, + "mean_feedback_forward_cosine": 0.9437725793231617, + "mean_feedback_forward_relative_error": 0.3248328838836063, + "min_feedback_forward_cosine": 0.8820503354072571 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026809988737441707, + "instruction_rms": 0.0026809988743536637, + "raw_apical_rms": 0.15930991488658539, + "teaching_rms": 0.0026809988737441707, + "traffic_rms": 0.15928673144668157 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.451622418055194e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.499655439701651e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017403166451312478, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 0.23905894103050232 + }, + { + "epoch": 41, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48141470551490784, + "mean_feedback_forward_cosine": 0.9476395476948131, + "mean_feedback_forward_relative_error": 0.31325085881081494, + "min_feedback_forward_cosine": 0.8889108896255493 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026126648043088025, + "instruction_rms": 0.0026126648047255247, + "raw_apical_rms": 0.15997367415926766, + "teaching_rms": 0.0026126648043088025, + "traffic_rms": 0.1599517542668094 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.997163547519449e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5042260491372747e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016357657258046936, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 0.2279548183441162 + }, + { + "epoch": 42, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4649565815925598, + "mean_feedback_forward_cosine": 0.9515861738811839, + "mean_feedback_forward_relative_error": 0.3011077326807109, + "min_feedback_forward_cosine": 0.8962357044219971 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026664416229715475, + "instruction_rms": 0.0026664416229580483, + "raw_apical_rms": 0.16075113269946528, + "teaching_rms": 0.0026664416229715475, + "traffic_rms": 0.16072843122039027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0866227029282527e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.495727531765114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01690388085456133, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 0.2309040823698044 + }, + { + "epoch": 43, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45108336210250854, + "mean_feedback_forward_cosine": 0.9550073363564231, + "mean_feedback_forward_relative_error": 0.2901201928203756, + "min_feedback_forward_cosine": 0.9022436141967773 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002620362023628201, + "instruction_rms": 0.0026203620228770833, + "raw_apical_rms": 0.16127809903963275, + "teaching_rms": 0.002620362023628201, + "traffic_rms": 0.16125625139461272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.0269003070106919e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.499343878914618e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016598557232337732, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 0.22040659890174866 + }, + { + "epoch": 44, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4375098645687103, + "mean_feedback_forward_cosine": 0.9582249500534751, + "mean_feedback_forward_relative_error": 0.27943304668773306, + "min_feedback_forward_cosine": 0.9080528020858765 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025907868207275977, + "instruction_rms": 0.0025907868211575463, + "raw_apical_rms": 0.16220559519125596, + "teaching_rms": 0.0025907868207275977, + "traffic_rms": 0.16218430218432747 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1013457879016642e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5082682182462002e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01642502517580754, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 0.2202847241783142 + }, + { + "epoch": 45, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4238121509552002, + "mean_feedback_forward_cosine": 0.9612270246852528, + "mean_feedback_forward_relative_error": 0.2690422976558859, + "min_feedback_forward_cosine": 0.9134528636932373 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002619014921180047, + "instruction_rms": 0.002619014920820418, + "raw_apical_rms": 0.1630859194538741, + "teaching_rms": 0.002619014921180047, + "traffic_rms": 0.16306428153431496 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 8.699811360202148e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4940739896583085e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017102781590817837, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 0.2182222050189972 + }, + { + "epoch": 46, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40952426195144653, + "mean_feedback_forward_cosine": 0.9640638784928756, + "mean_feedback_forward_relative_error": 0.2589003778316758, + "min_feedback_forward_cosine": 0.9190661311149597 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026306900886823334, + "instruction_rms": 0.0026306900890591266, + "raw_apical_rms": 0.16387086393761952, + "teaching_rms": 0.0026306900886823334, + "traffic_rms": 0.1638492099552594 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.2537168458948145e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4975970813183435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016335139995428287, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 0.2186380574798584 + }, + { + "epoch": 47, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3954333961009979, + "mean_feedback_forward_cosine": 0.9666750344363126, + "mean_feedback_forward_relative_error": 0.24921436350453985, + "min_feedback_forward_cosine": 0.9244191646575928 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026093564115870654, + "instruction_rms": 0.0026093564112593093, + "raw_apical_rms": 0.16429322970740004, + "teaching_rms": 0.0026093564115870654, + "traffic_rms": 0.1642719300649487 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.677049206402444e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.495478714290304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016260944054722705, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 0.20563692422389984 + }, + { + "epoch": 48, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38231518864631653, + "mean_feedback_forward_cosine": 0.9691955317150462, + "mean_feedback_forward_relative_error": 0.2395152908834544, + "min_feedback_forward_cosine": 0.929153323173523 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026364211260341648, + "instruction_rms": 0.002636421126467876, + "raw_apical_rms": 0.1644946603299624, + "teaching_rms": 0.0026364211260341648, + "traffic_rms": 0.1644729018117333 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0652374271558074e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4934711062164372e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016440977500770212, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 0.2118056460905075 + }, + { + "epoch": 49, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36732879281044006, + "mean_feedback_forward_cosine": 0.9715332399715076, + "mean_feedback_forward_relative_error": 0.23023826757615262, + "min_feedback_forward_cosine": 0.9346952438354492 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00261836602578771, + "instruction_rms": 0.002618366026175186, + "raw_apical_rms": 0.16503011495313352, + "teaching_rms": 0.00261836602578771, + "traffic_rms": 0.16500874538808866 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.1025967694422434e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4937622900376938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016367926852218582, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 0.20707733969211578 + }, + { + "epoch": 50, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35372257232666016, + "mean_feedback_forward_cosine": 0.9736507556655191, + "mean_feedback_forward_relative_error": 0.22146562026305633, + "min_feedback_forward_cosine": 0.939509391784668 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026116874379210576, + "instruction_rms": 0.0026116874380828154, + "raw_apical_rms": 0.1652809145581337, + "teaching_rms": 0.0026116874379210576, + "traffic_rms": 0.16525966183780944 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0346428069851754e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4924827854771114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016596063956498347, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 0.20197061972618102 + }, + { + "epoch": 51, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3421255350112915, + "mean_feedback_forward_cosine": 0.975636520168998, + "mean_feedback_forward_relative_error": 0.21288804323835805, + "min_feedback_forward_cosine": 0.9432780146598816 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026388746050884817, + "instruction_rms": 0.0026388746049235945, + "raw_apical_rms": 0.16652388719459782, + "teaching_rms": 0.0026388746050884817, + "traffic_rms": 0.16650233102139708 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0353338808499757e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.492350495358787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017267594974589004, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 0.20091737973213195 + }, + { + "epoch": 52, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33032992482185364, + "mean_feedback_forward_cosine": 0.9773970159617338, + "mean_feedback_forward_relative_error": 0.20498079054734925, + "min_feedback_forward_cosine": 0.947166919708252 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025517153861885043, + "instruction_rms": 0.002551715386164629, + "raw_apical_rms": 0.166553919412207, + "teaching_rms": 0.0025517153861885043, + "traffic_rms": 0.16653380855515165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0046150578091328e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4905899280037668e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016598467752594662, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 0.19553858379840852 + }, + { + "epoch": 53, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3186520040035248, + "mean_feedback_forward_cosine": 0.9790833083066073, + "mean_feedback_forward_relative_error": 0.1971417141908949, + "min_feedback_forward_cosine": 0.9506666660308838 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026076348953039848, + "instruction_rms": 0.002607634895725366, + "raw_apical_rms": 0.16719893172432032, + "teaching_rms": 0.0026076348953039848, + "traffic_rms": 0.16717794182583393 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.322648728319109e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4923480867364188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016862344446026238, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 0.19238390841722489 + }, + { + "epoch": 54, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30720165371894836, + "mean_feedback_forward_cosine": 0.9806832421909679, + "mean_feedback_forward_relative_error": 0.18942306989973243, + "min_feedback_forward_cosine": 0.9541330337524414 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002651756099626751, + "instruction_rms": 0.0026517560999571115, + "raw_apical_rms": 0.16776643555320603, + "teaching_rms": 0.002651756099626751, + "traffic_rms": 0.16774485174847997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.003275636132457e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4936352147016205e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01776093132442331, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 0.19021243114948272 + }, + { + "epoch": 55, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29608866572380066, + "mean_feedback_forward_cosine": 0.9821706067432057, + "mean_feedback_forward_relative_error": 0.18196624409068715, + "min_feedback_forward_cosine": 0.9573959112167358 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002637902870904746, + "instruction_rms": 0.002637902870982881, + "raw_apical_rms": 0.16845634147371846, + "teaching_rms": 0.002637902870904746, + "traffic_rms": 0.16843500418973295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.180890265572998e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4929516753856554e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016787107435931564, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 0.19022533592224122 + }, + { + "epoch": 56, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28472211956977844, + "mean_feedback_forward_cosine": 0.9835448102517561, + "mean_feedback_forward_relative_error": 0.17479036572304638, + "min_feedback_forward_cosine": 0.9606002569198608 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002640355089161235, + "instruction_rms": 0.0026403550885563142, + "raw_apical_rms": 0.16918120292000444, + "teaching_rms": 0.002640355089161235, + "traffic_rms": 0.16915984221272448 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546024680137634, + "maximum_absolute_post_projection_soma_slope": 1.0484234280738747e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4888118480348777e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016646698143046222, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 0.1905882199382782 + }, + { + "epoch": 57, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2741278111934662, + "mean_feedback_forward_cosine": 0.9848274566910483, + "mean_feedback_forward_relative_error": 0.1678287715397098, + "min_feedback_forward_cosine": 0.9634441137313843 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002645610374618366, + "instruction_rms": 0.002645610374570658, + "raw_apical_rms": 0.16984846648163496, + "teaching_rms": 0.002645610374618366, + "traffic_rms": 0.16982713521475845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.740586648376848e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4892988633041292e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016189196600952662, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 0.19333840260982513 + }, + { + "epoch": 58, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26488324999809265, + "mean_feedback_forward_cosine": 0.985942682352933, + "mean_feedback_forward_relative_error": 0.16151486520062794, + "min_feedback_forward_cosine": 0.965793251991272 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025522345421821716, + "instruction_rms": 0.002552234542252195, + "raw_apical_rms": 0.1706680666336003, + "teaching_rms": 0.0025522345421821716, + "traffic_rms": 0.17064832860810458 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0020428220514077e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4825187181228823e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016579213444880603, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 0.18492080654382706 + }, + { + "epoch": 59, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25429821014404297, + "mean_feedback_forward_cosine": 0.9870492946017873, + "mean_feedback_forward_relative_error": 0.1550373097035018, + "min_feedback_forward_cosine": 0.9684604406356812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002638056296315936, + "instruction_rms": 0.002638056296902102, + "raw_apical_rms": 0.17076245447727467, + "teaching_rms": 0.002638056296315936, + "traffic_rms": 0.17074145527113044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1674350730572769e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.482717339769628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01650908573897969, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 0.18232514607429504 + }, + { + "epoch": 60, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2450953722000122, + "mean_feedback_forward_cosine": 0.9879905007102273, + "mean_feedback_forward_relative_error": 0.14927273683927275, + "min_feedback_forward_cosine": 0.9706327319145203 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002550035057703783, + "instruction_rms": 0.0025500350577966416, + "raw_apical_rms": 0.1712080569415117, + "teaching_rms": 0.002550035057703783, + "traffic_rms": 0.17118849477339904 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0393012672693658e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4843021173125663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016497432091448293, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 0.17654647376060487 + }, + { + "epoch": 61, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23602604866027832, + "mean_feedback_forward_cosine": 0.9888945406133478, + "mean_feedback_forward_relative_error": 0.14352597526528618, + "min_feedback_forward_cosine": 0.9726927280426025 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025810403677181057, + "instruction_rms": 0.0025810403672865004, + "raw_apical_rms": 0.17187053374015993, + "teaching_rms": 0.0025810403677181057, + "traffic_rms": 0.1718504733660613 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 8.796149586487445e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4847600552706405e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016735673653844367, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 0.17747351435661315 + }, + { + "epoch": 62, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22772634029388428, + "mean_feedback_forward_cosine": 0.9897308739748868, + "mean_feedback_forward_relative_error": 0.13800441101193428, + "min_feedback_forward_cosine": 0.9746429920196533 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002586567051918204, + "instruction_rms": 0.0025865670514441204, + "raw_apical_rms": 0.17275806818638229, + "teaching_rms": 0.002586567051918204, + "traffic_rms": 0.17273807663806964 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0840110320486929e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479037982036916e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01650347677460205, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 0.18059674732208253 + }, + { + "epoch": 63, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2193385362625122, + "mean_feedback_forward_cosine": 0.9905158530582081, + "mean_feedback_forward_relative_error": 0.13260736763477327, + "min_feedback_forward_cosine": 0.9764289855957031 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002593862720361383, + "instruction_rms": 0.002593862720415786, + "raw_apical_rms": 0.17355899855922224, + "teaching_rms": 0.002593862720361383, + "traffic_rms": 0.17353897890513706 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0346015955065013e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.489791023663665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01763454322460327, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 0.18066855516433716 + }, + { + "epoch": 64, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21023254096508026, + "mean_feedback_forward_cosine": 0.991244206645272, + "mean_feedback_forward_relative_error": 0.12742867950688708, + "min_feedback_forward_cosine": 0.978335976600647 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025885607366504877, + "instruction_rms": 0.0025885607366363844, + "raw_apical_rms": 0.17384061877452603, + "teaching_rms": 0.0025885607366504877, + "traffic_rms": 0.17382071288685452 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 8.765253056708389e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484657846088851e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017558244499005327, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 0.18013090616226196 + }, + { + "epoch": 65, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20314325392246246, + "mean_feedback_forward_cosine": 0.9918967745520851, + "mean_feedback_forward_relative_error": 0.12256507690657269, + "min_feedback_forward_cosine": 0.9797675609588623 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025573996431796555, + "instruction_rms": 0.0025573996428686894, + "raw_apical_rms": 0.17463030524648776, + "teaching_rms": 0.0025573996431796555, + "traffic_rms": 0.17461095065712454 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.553939150919177e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484204877039705e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017377423911661544, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 0.1738362654709816 + }, + { + "epoch": 66, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19643403589725494, + "mean_feedback_forward_cosine": 0.9924948681484569, + "mean_feedback_forward_relative_error": 0.11792191006243229, + "min_feedback_forward_cosine": 0.9810354709625244 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025724413201511326, + "instruction_rms": 0.0025724413209525245, + "raw_apical_rms": 0.17510939914803397, + "teaching_rms": 0.0025724413201511326, + "traffic_rms": 0.17508981700897086 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.099477913157898e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.481592396424525e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015943953697485098, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 0.16761100320339203 + }, + { + "epoch": 67, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18974581360816956, + "mean_feedback_forward_cosine": 0.9930525736375289, + "mean_feedback_forward_relative_error": 0.11344104253432967, + "min_feedback_forward_cosine": 0.9823153018951416 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002562874594663576, + "instruction_rms": 0.002562874594672658, + "raw_apical_rms": 0.17572361243482387, + "teaching_rms": 0.002562874594663576, + "traffic_rms": 0.17570437209766812 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.155368019264642e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4836223681102065e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017682588871554217, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 0.1711928145122528 + }, + { + "epoch": 68, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18149156868457794, + "mean_feedback_forward_cosine": 0.993578559702093, + "mean_feedback_forward_relative_error": 0.10905899239534682, + "min_feedback_forward_cosine": 0.9837833642959595 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002573964630008046, + "instruction_rms": 0.0025739646302974646, + "raw_apical_rms": 0.17695228248226894, + "teaching_rms": 0.002573964630008046, + "traffic_rms": 0.1769328336222889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.345085061340797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.481487556816963e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01638943024393122, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 0.16589756311416626 + }, + { + "epoch": 69, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17431750893592834, + "mean_feedback_forward_cosine": 0.9940657648173246, + "mean_feedback_forward_relative_error": 0.10483911863782189, + "min_feedback_forward_cosine": 0.9850492477416992 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025773425199029365, + "instruction_rms": 0.0025773425208361405, + "raw_apical_rms": 0.17804094700946912, + "teaching_rms": 0.0025773425199029365, + "traffic_rms": 0.17802166814134152 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 8.992429911813815e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.486506641525877e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01760795812856261, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 0.1724113479423523 + }, + { + "epoch": 70, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16705475747585297, + "mean_feedback_forward_cosine": 0.9945228901776401, + "mean_feedback_forward_relative_error": 0.10072282813489437, + "min_feedback_forward_cosine": 0.986244797706604 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002582189546796684, + "instruction_rms": 0.0025821895468024388, + "raw_apical_rms": 0.1787770083718866, + "teaching_rms": 0.002582189546796684, + "traffic_rms": 0.17875768991575292 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0272326989024805e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4775851635873862e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01686638128413306, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 0.16539470848083496 + }, + { + "epoch": 71, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1604367047548294, + "mean_feedback_forward_cosine": 0.9949283513155851, + "mean_feedback_forward_relative_error": 0.09693231247365475, + "min_feedback_forward_cosine": 0.9873062372207642 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025467370097468586, + "instruction_rms": 0.0025467370101966533, + "raw_apical_rms": 0.17909838920990148, + "teaching_rms": 0.0025467370097468586, + "traffic_rms": 0.17907960410402332 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.973010861585863e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4817502037248477e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01745220751057537, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 0.1582150242996216 + }, + { + "epoch": 72, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1535358428955078, + "mean_feedback_forward_cosine": 0.9953303911469199, + "mean_feedback_forward_relative_error": 0.09303388673473488, + "min_feedback_forward_cosine": 0.988383412361145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025991587708195686, + "instruction_rms": 0.0025991587716081913, + "raw_apical_rms": 0.18000763324335597, + "teaching_rms": 0.0025991587708195686, + "traffic_rms": 0.17998812637682537 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0352358259524408e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4849418622237236e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017115589551013798, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 0.16939524397850036 + }, + { + "epoch": 73, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14759239554405212, + "mean_feedback_forward_cosine": 0.9956857345320962, + "mean_feedback_forward_relative_error": 0.08942185426977547, + "min_feedback_forward_cosine": 0.9892558455467224 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025794316914105123, + "instruction_rms": 0.0025794316913048793, + "raw_apical_rms": 0.18022014019060226, + "teaching_rms": 0.0025794316914105123, + "traffic_rms": 0.1802009989696813 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.445389537177107e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4819788111199093e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017486633000842515, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 0.163170988407135 + }, + { + "epoch": 74, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1414351463317871, + "mean_feedback_forward_cosine": 0.9960176251151345, + "mean_feedback_forward_relative_error": 0.08593806465241043, + "min_feedback_forward_cosine": 0.9901398420333862 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00256953132857665, + "instruction_rms": 0.002569531328551776, + "raw_apical_rms": 0.18083494651140056, + "teaching_rms": 0.00256953132857665, + "traffic_rms": 0.18081604357852052 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.852307414348616e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4809725515344927e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016785896326763583, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 0.16299445527553558 + }, + { + "epoch": 75, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13638880848884583, + "mean_feedback_forward_cosine": 0.9963149428367615, + "mean_feedback_forward_relative_error": 0.08266138681633906, + "min_feedback_forward_cosine": 0.9908096194267273 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002552928812818154, + "instruction_rms": 0.0025529288124809322, + "raw_apical_rms": 0.18192648726296556, + "teaching_rms": 0.002552928812818154, + "traffic_rms": 0.1819078983262884 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0283193319082784e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480736704792101e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016662293143641125, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 0.1621455531024933 + }, + { + "epoch": 76, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13116276264190674, + "mean_feedback_forward_cosine": 0.996587411923842, + "mean_feedback_forward_relative_error": 0.07954189191487702, + "min_feedback_forward_cosine": 0.9915049076080322 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025371212997851774, + "instruction_rms": 0.0025371213000250376, + "raw_apical_rms": 0.18197764352736884, + "teaching_rms": 0.0025371212997851774, + "traffic_rms": 0.1819592364754869 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0026366226156824e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4799739870315653e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015684972572630386, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 0.15983858047485353 + }, + { + "epoch": 77, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12551027536392212, + "mean_feedback_forward_cosine": 0.9968422813849016, + "mean_feedback_forward_relative_error": 0.0765163877132264, + "min_feedback_forward_cosine": 0.992214560508728 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025360878619202763, + "instruction_rms": 0.0025360878624729427, + "raw_apical_rms": 0.18284192726495202, + "teaching_rms": 0.0025360878619202763, + "traffic_rms": 0.18282354382644486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.493937369597916e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4788402425795733e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016325941678016397, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 0.15839524523258208 + }, + { + "epoch": 78, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12059476226568222, + "mean_feedback_forward_cosine": 0.9970832900567488, + "mean_feedback_forward_relative_error": 0.07354833676733753, + "min_feedback_forward_cosine": 0.9928189516067505 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025638492129064134, + "instruction_rms": 0.0025638492124394254, + "raw_apical_rms": 0.1837232627577498, + "teaching_rms": 0.0025638492129064134, + "traffic_rms": 0.183704718435586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0931784544254697e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4804358192345185e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01753654451022943, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 0.15468290334701537 + }, + { + "epoch": 79, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1153392344713211, + "mean_feedback_forward_cosine": 0.9973155617713928, + "mean_feedback_forward_relative_error": 0.07057084146548401, + "min_feedback_forward_cosine": 0.9934232234954834 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026191846587147035, + "instruction_rms": 0.002619184658812227, + "raw_apical_rms": 0.1839992865564056, + "teaching_rms": 0.0026191846587147035, + "traffic_rms": 0.1839799958113869 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.47943377127558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4868803208244217e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01905084658051881, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 0.15875575710773468 + }, + { + "epoch": 80, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11110249161720276, + "mean_feedback_forward_cosine": 0.9975191192193464, + "mean_feedback_forward_relative_error": 0.06784528554840522, + "min_feedback_forward_cosine": 0.993894636631012 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025777914235396663, + "instruction_rms": 0.0025777914236338015, + "raw_apical_rms": 0.18410893212336307, + "teaching_rms": 0.0025777914235396663, + "traffic_rms": 0.18409022851033302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0275822148742009e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4807668011633497e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018296956831000064, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 0.15650936687231065 + }, + { + "epoch": 81, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10675564408302307, + "mean_feedback_forward_cosine": 0.9977120724591342, + "mean_feedback_forward_relative_error": 0.06514448926530102, + "min_feedback_forward_cosine": 0.9943588972091675 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002607873224825701, + "instruction_rms": 0.0026078732248203482, + "raw_apical_rms": 0.1846595213252752, + "teaching_rms": 0.002607873224825701, + "traffic_rms": 0.18464036516317262 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1198822846836265e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4794337226069873e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01678601461984566, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 0.1572362718772888 + }, + { + "epoch": 82, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10299966484308243, + "mean_feedback_forward_cosine": 0.9978804425759749, + "mean_feedback_forward_relative_error": 0.06270305947823958, + "min_feedback_forward_cosine": 0.9947507381439209 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002527017969359533, + "instruction_rms": 0.002527017969319647, + "raw_apical_rms": 0.1857704175019096, + "teaching_rms": 0.002527017969359533, + "traffic_rms": 0.18575255227407314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.573723502853682e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4797205407453714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018375971280391173, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 0.15369563507556916 + }, + { + "epoch": 83, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09909442067146301, + "mean_feedback_forward_cosine": 0.998041033744812, + "mean_feedback_forward_relative_error": 0.06028071604669094, + "min_feedback_forward_cosine": 0.9951369762420654 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025772767509259728, + "instruction_rms": 0.002577276750747814, + "raw_apical_rms": 0.1864720889989372, + "teaching_rms": 0.0025772767509259728, + "traffic_rms": 0.18645351399129914 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0927622184908614e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4828781723599312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016239503241216927, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 0.15495589464187623 + }, + { + "epoch": 84, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0953102633357048, + "mean_feedback_forward_cosine": 0.9981857754967429, + "mean_feedback_forward_relative_error": 0.05801168021491983, + "min_feedback_forward_cosine": 0.9954981207847595 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025373207589904764, + "instruction_rms": 0.002537320759036161, + "raw_apical_rms": 0.18698361611123976, + "teaching_rms": 0.0025373207589904764, + "traffic_rms": 0.18696575609022667 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.4442558438076958e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4841613851061705e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016575539687734793, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 0.14936254230499268 + }, + { + "epoch": 85, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09176845103502274, + "mean_feedback_forward_cosine": 0.9983182365244085, + "mean_feedback_forward_relative_error": 0.05585403787818822, + "min_feedback_forward_cosine": 0.9958264827728271 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002528046100384672, + "instruction_rms": 0.002528046100279032, + "raw_apical_rms": 0.18789531201300305, + "teaching_rms": 0.002528046100384672, + "traffic_rms": 0.18787758940576382 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.2625748979644413e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4873195438157574e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0175174926644487, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 0.14558772820949553 + }, + { + "epoch": 86, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08846888691186905, + "mean_feedback_forward_cosine": 0.998448104208166, + "mean_feedback_forward_relative_error": 0.053648963028734384, + "min_feedback_forward_cosine": 0.9961194396018982 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00259800201447266, + "instruction_rms": 0.0025980020140611007, + "raw_apical_rms": 0.18825787565019297, + "teaching_rms": 0.00259800201447266, + "traffic_rms": 0.1882391808774175 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0122546711954783e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4816388196931944e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017309788337355355, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 0.1539412923526764 + }, + { + "epoch": 87, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08506879210472107, + "mean_feedback_forward_cosine": 0.9985643711957064, + "mean_feedback_forward_relative_error": 0.051602129909125244, + "min_feedback_forward_cosine": 0.996408224105835 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002564680610138021, + "instruction_rms": 0.0025646806100824066, + "raw_apical_rms": 0.1882102230749297, + "teaching_rms": 0.002564680610138021, + "traffic_rms": 0.18819208920688474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0245884851656228e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4800534997113026e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.019056134583151776, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 0.1452641355419159 + }, + { + "epoch": 88, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08190954476594925, + "mean_feedback_forward_cosine": 0.9986741228537126, + "mean_feedback_forward_relative_error": 0.049598096700554546, + "min_feedback_forward_cosine": 0.9966740608215332 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025392887658598653, + "instruction_rms": 0.0025392887656967194, + "raw_apical_rms": 0.18920585043068994, + "teaching_rms": 0.0025392887658598653, + "traffic_rms": 0.18918811979316316 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.839197190331106e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4887311440596065e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016645079352624978, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 0.14693725987911224 + }, + { + "epoch": 89, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0789809376001358, + "mean_feedback_forward_cosine": 0.9987735098058527, + "mean_feedback_forward_relative_error": 0.04770510923117399, + "min_feedback_forward_cosine": 0.9969068169593811 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025462232352789545, + "instruction_rms": 0.002546223234908394, + "raw_apical_rms": 0.18932612261344742, + "teaching_rms": 0.0025462232352789545, + "traffic_rms": 0.18930825714827643 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.246823395869797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.484017046954309e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015969287259243948, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 0.14764454377174377 + }, + { + "epoch": 90, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07583211362361908, + "mean_feedback_forward_cosine": 0.9988703196698969, + "mean_feedback_forward_relative_error": 0.04578968093476512, + "min_feedback_forward_cosine": 0.9971469044685364 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026005807915421356, + "instruction_rms": 0.0026005807911986165, + "raw_apical_rms": 0.1899842957352843, + "teaching_rms": 0.0026005807915421356, + "traffic_rms": 0.18996581956085978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.68694067182696e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4833008125752398e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01733682473716961, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 0.1446842994451523 + }, + { + "epoch": 91, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07291347533464432, + "mean_feedback_forward_cosine": 0.9989564884792674, + "mean_feedback_forward_relative_error": 0.04401236115531488, + "min_feedback_forward_cosine": 0.9973591566085815 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002576481053850996, + "instruction_rms": 0.0025764810536133866, + "raw_apical_rms": 0.19055341867911108, + "teaching_rms": 0.002576481053850996, + "traffic_rms": 0.19053516427050277 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0505418401862698e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4801907337077965e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017994937508858715, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 0.1430196456503868 + }, + { + "epoch": 92, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06965237855911255, + "mean_feedback_forward_cosine": 0.9990409677678889, + "mean_feedback_forward_relative_error": 0.04220309384505857, + "min_feedback_forward_cosine": 0.9975884556770325 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002628571878665227, + "instruction_rms": 0.0026285718789185474, + "raw_apical_rms": 0.19078388602103827, + "teaching_rms": 0.002628571878665227, + "traffic_rms": 0.19076502704289183 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0332728095363564e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4831468182744168e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01653179728919613, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 0.15168742237091065 + }, + { + "epoch": 93, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06699279695749283, + "mean_feedback_forward_cosine": 0.9991126103834672, + "mean_feedback_forward_relative_error": 0.04059575887566263, + "min_feedback_forward_cosine": 0.9977678060531616 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025268405760705483, + "instruction_rms": 0.0025268405756366614, + "raw_apical_rms": 0.19115784172048894, + "teaching_rms": 0.0025268405760705483, + "traffic_rms": 0.19114040081660344 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.432229575168094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4835569481579387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016920708942891202, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 0.14105878267526625 + }, + { + "epoch": 94, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06481534242630005, + "mean_feedback_forward_cosine": 0.9991770950230685, + "mean_feedback_forward_relative_error": 0.0390908916565505, + "min_feedback_forward_cosine": 0.9979104995727539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002531399552439272, + "instruction_rms": 0.00253139955218531, + "raw_apical_rms": 0.19195411250876204, + "teaching_rms": 0.002531399552439272, + "traffic_rms": 0.19193682479920285 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0476949086068998e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4844427994187307e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015994308793579683, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 0.14001432754039764 + }, + { + "epoch": 95, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06230797618627548, + "mean_feedback_forward_cosine": 0.9992411147464405, + "mean_feedback_forward_relative_error": 0.03754272110421549, + "min_feedback_forward_cosine": 0.9980698227882385 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002586975071940399, + "instruction_rms": 0.002586975071685724, + "raw_apical_rms": 0.19217737725612175, + "teaching_rms": 0.002586975071940399, + "traffic_rms": 0.1921592560378314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.142328849520709e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.487447375191788e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01634697191483834, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 0.1472200103521347 + }, + { + "epoch": 96, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059814244508743286, + "mean_feedback_forward_cosine": 0.9993016600608826, + "mean_feedback_forward_relative_error": 0.036020663820884445, + "min_feedback_forward_cosine": 0.9982219934463501 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026100477892132164, + "instruction_rms": 0.002610047788790033, + "raw_apical_rms": 0.19195360327957614, + "teaching_rms": 0.0026100477892132164, + "traffic_rms": 0.19193515086242044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.352968532994055e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4856347337514025e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016519331770788452, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 0.14532155244827272 + }, + { + "epoch": 97, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05754891782999039, + "mean_feedback_forward_cosine": 0.9993545857342807, + "mean_feedback_forward_relative_error": 0.034629703922705216, + "min_feedback_forward_cosine": 0.9983534216880798 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002576737762767766, + "instruction_rms": 0.002576737762973315, + "raw_apical_rms": 0.1923760315450754, + "teaching_rms": 0.002576737762767766, + "traffic_rms": 0.1923581634292759 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.057631444875369e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.48435047300443e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016919312137844044, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 0.14420203866481782 + }, + { + "epoch": 98, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05534670129418373, + "mean_feedback_forward_cosine": 0.9994056593288075, + "mean_feedback_forward_relative_error": 0.03323356888510964, + "min_feedback_forward_cosine": 0.9984756708145142 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026261368710812213, + "instruction_rms": 0.0026261368709735067, + "raw_apical_rms": 0.1930536316891598, + "teaching_rms": 0.0026261368710812213, + "traffic_rms": 0.1930349421822663 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1575757241644169e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4865180520205188e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017523810182584847, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 0.14944930130958556 + }, + { + "epoch": 99, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05311473459005356, + "mean_feedback_forward_cosine": 0.9994519504633816, + "mean_feedback_forward_relative_error": 0.03191699786958369, + "min_feedback_forward_cosine": 0.9985951781272888 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025835425225691484, + "instruction_rms": 0.0025835425220691226, + "raw_apical_rms": 0.19418834360437803, + "teaching_rms": 0.0025835425225691484, + "traffic_rms": 0.19417054309543438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0065055988661697e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4848001618888635e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.018626418220019314, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 0.1464569148015976 + }, + { + "epoch": 100, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05120520293712616, + "mean_feedback_forward_cosine": 0.9994918682358481, + "mean_feedback_forward_relative_error": 0.03073248685422269, + "min_feedback_forward_cosine": 0.9986938834190369 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025275149174151176, + "instruction_rms": 0.0025275149175734177, + "raw_apical_rms": 0.19497533635921768, + "teaching_rms": 0.0025275149174151176, + "traffic_rms": 0.19495830988015658 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1913288489040497e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480554247544735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01760787833682661, + "minimum_observations": 80 + }, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 0.1350346232509613 + }, + { + "epoch": 101, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05119210481643677, + "mean_feedback_forward_cosine": 0.9994921814311635, + "mean_feedback_forward_relative_error": 0.030721981671046127, + "min_feedback_forward_cosine": 0.9986945390701294 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0017687354392080935, + "instruction_rms": 0.001768735439418279, + "raw_apical_rms": 0.19499570977986463, + "teaching_rms": 0.0017687354392080935, + "traffic_rms": 0.19498698580837653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1586549675257629e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474734706299098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016416850248251326, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 0.07555322633504867 + }, + { + "epoch": 102, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05118563026189804, + "mean_feedback_forward_cosine": 0.9994923353195191, + "mean_feedback_forward_relative_error": 0.030716231143610044, + "min_feedback_forward_cosine": 0.998694896697998 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014938466325539568, + "instruction_rms": 0.0014938466323345856, + "raw_apical_rms": 0.19499805159988684, + "teaching_rms": 0.0014938466325539568, + "traffic_rms": 0.19499176047063133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0710802911262363e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.480583964358247e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016293858728893787, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 0.049803137357234954 + }, + { + "epoch": 103, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051180191338062286, + "mean_feedback_forward_cosine": 0.9994924642822959, + "mean_feedback_forward_relative_error": 0.030711198259483683, + "min_feedback_forward_cosine": 0.9986951947212219 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001395310569594246, + "instruction_rms": 0.0013953105693609211, + "raw_apical_rms": 0.19497426753524605, + "teaching_rms": 0.001395310569594246, + "traffic_rms": 0.19496869230617547 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0260776406312289e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.471476376770101e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01601534201279273, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 0.039833071542978285 + }, + { + "epoch": 104, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05117492005228996, + "mean_feedback_forward_cosine": 0.9994925964962352, + "mean_feedback_forward_relative_error": 0.030706490643999794, + "min_feedback_forward_cosine": 0.998695433139801 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013467496702160577, + "instruction_rms": 0.001346749670120203, + "raw_apical_rms": 0.19493982587143346, + "teaching_rms": 0.0013467496702160577, + "traffic_rms": 0.1949345462200083 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.607982320858355e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.470372273502311e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015758368506259134, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 0.035190497146844864 + }, + { + "epoch": 105, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051169347018003464, + "mean_feedback_forward_cosine": 0.9994927178729665, + "mean_feedback_forward_relative_error": 0.03070180763236501, + "min_feedback_forward_cosine": 0.9986957311630249 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013410266706276402, + "instruction_rms": 0.0013410266704459784, + "raw_apical_rms": 0.1949389816113546, + "teaching_rms": 0.0013410266706276402, + "traffic_rms": 0.19493359573060845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 8.470676249316966e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4739950526730677e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015929924881889868, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 0.03185617652177811 + }, + { + "epoch": 106, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05116403102874756, + "mean_feedback_forward_cosine": 0.9994928490031849, + "mean_feedback_forward_relative_error": 0.030697416607290508, + "min_feedback_forward_cosine": 0.9986958503723145 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012698655848057498, + "instruction_rms": 0.0012698655848547207, + "raw_apical_rms": 0.19492830158761082, + "teaching_rms": 0.0012698655848057498, + "traffic_rms": 0.19492347456669973 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.682968737934061e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4777255905422743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01634822931766062, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 0.027642367269992828 + }, + { + "epoch": 107, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051158614456653595, + "mean_feedback_forward_cosine": 0.9994929909706116, + "mean_feedback_forward_relative_error": 0.030693123938346452, + "min_feedback_forward_cosine": 0.9986962080001831 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012322417702421802, + "instruction_rms": 0.001232241770020053, + "raw_apical_rms": 0.1948894802009726, + "teaching_rms": 0.0012322417702421802, + "traffic_rms": 0.19488473887491367 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.77276926050763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.474710634322149e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01637651106613543, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 0.02568652232527733 + }, + { + "epoch": 108, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05115358158946037, + "mean_feedback_forward_cosine": 0.9994930863380432, + "mean_feedback_forward_relative_error": 0.030688967191698877, + "min_feedback_forward_cosine": 0.9986964464187622 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001226166680269132, + "instruction_rms": 0.0012261666802818028, + "raw_apical_rms": 0.19485631247614754, + "teaching_rms": 0.001226166680269132, + "traffic_rms": 0.19485169718657128 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.755269303468594e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4753380043331286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016056361506477074, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 0.023811488244831563 + }, + { + "epoch": 109, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05114837735891342, + "mean_feedback_forward_cosine": 0.9994931947101247, + "mean_feedback_forward_relative_error": 0.030684860694137485, + "min_feedback_forward_cosine": 0.9986968040466309 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012075743669640346, + "instruction_rms": 0.0012075743670406517, + "raw_apical_rms": 0.1948191290338282, + "teaching_rms": 0.0012075743669640346, + "traffic_rms": 0.1948145974231042 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.437042081117397e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479210040453366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016448864817390725, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 0.021838345532417296 + }, + { + "epoch": 110, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05114351585507393, + "mean_feedback_forward_cosine": 0.999493332342668, + "mean_feedback_forward_relative_error": 0.03068094899539243, + "min_feedback_forward_cosine": 0.9986969828605652 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011595853949311966, + "instruction_rms": 0.0011595853948911205, + "raw_apical_rms": 0.194762544911269, + "teaching_rms": 0.0011595853949311966, + "traffic_rms": 0.19475825461206062 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0166128561195364e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4763900455889074e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01631016846317762, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 0.019726173406243323 + }, + { + "epoch": 111, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05113827437162399, + "mean_feedback_forward_cosine": 0.999493437463587, + "mean_feedback_forward_relative_error": 0.030676899257708678, + "min_feedback_forward_cosine": 0.9986971616744995 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011874593887292162, + "instruction_rms": 0.001187459388700077, + "raw_apical_rms": 0.19477418710483002, + "teaching_rms": 0.0011874593887292162, + "traffic_rms": 0.19476971495636214 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0195687139002985e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4844790596748583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016586395925467047, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 0.01932826031059027 + }, + { + "epoch": 112, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051133524626493454, + "mean_feedback_forward_cosine": 0.9994935393333435, + "mean_feedback_forward_relative_error": 0.03067325239323757, + "min_feedback_forward_cosine": 0.9986975193023682 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010877423972940135, + "instruction_rms": 0.0010877423971210468, + "raw_apical_rms": 0.1947988182682167, + "teaching_rms": 0.0010877423972940135, + "traffic_rms": 0.19479480054823864 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.902134223693793e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4712276009725366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016392567859133407, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 0.017021890839487314 + }, + { + "epoch": 113, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051129020750522614, + "mean_feedback_forward_cosine": 0.9994936314496127, + "mean_feedback_forward_relative_error": 0.030669755166904494, + "min_feedback_forward_cosine": 0.9986977577209473 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001068225766504146, + "instruction_rms": 0.0010682257663737465, + "raw_apical_rms": 0.19472969165086776, + "teaching_rms": 0.001068225766504146, + "traffic_rms": 0.19472592476635744 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1057867510544384e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4778449875748578e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01683530956687084, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 0.015155086617395282 + }, + { + "epoch": 114, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05112415552139282, + "mean_feedback_forward_cosine": 0.9994937441565773, + "mean_feedback_forward_relative_error": 0.030666143100031396, + "min_feedback_forward_cosine": 0.9986979961395264 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010823955071153598, + "instruction_rms": 0.00108239550716524, + "raw_apical_rms": 0.1946883611477314, + "teaching_rms": 0.0010823955071153598, + "traffic_rms": 0.1946844195885319 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0401486605360333e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4731284545704562e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.017039823129188444, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 0.015156236096322537 + }, + { + "epoch": 115, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05111939087510109, + "mean_feedback_forward_cosine": 0.9994938492774963, + "mean_feedback_forward_relative_error": 0.030662513041699476, + "min_feedback_forward_cosine": 0.9986982941627502 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010619285310306747, + "instruction_rms": 0.0010619285310397503, + "raw_apical_rms": 0.19474934291508844, + "teaching_rms": 0.0010619285310306747, + "traffic_rms": 0.19474544012192432 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0616102485982992e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479102433570658e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01587878119676388, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 0.015015304269492626 + }, + { + "epoch": 116, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05111433565616608, + "mean_feedback_forward_cosine": 0.9994939587332986, + "mean_feedback_forward_relative_error": 0.030658745401623576, + "min_feedback_forward_cosine": 0.9986985921859741 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010654835011902415, + "instruction_rms": 0.0010654835013510324, + "raw_apical_rms": 0.19473436321227186, + "teaching_rms": 0.0010654835011902415, + "traffic_rms": 0.19473047003921887 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0532656347095326e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4729436117464805e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016423113784272993, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 0.01422905849918723 + }, + { + "epoch": 117, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05110981687903404, + "mean_feedback_forward_cosine": 0.9994940638542176, + "mean_feedback_forward_relative_error": 0.030655291176993737, + "min_feedback_forward_cosine": 0.9986987709999084 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010135628770398522, + "instruction_rms": 0.0010135628768825147, + "raw_apical_rms": 0.194692488239523, + "teaching_rms": 0.0010135628770398522, + "traffic_rms": 0.19468888714727084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0532649952210704e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47659023229905e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015894250125513886, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 0.013055190514028073 + }, + { + "epoch": 118, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05110520124435425, + "mean_feedback_forward_cosine": 0.9994941700588573, + "mean_feedback_forward_relative_error": 0.030651822203600947, + "min_feedback_forward_cosine": 0.9986990690231323 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000987230021848355, + "instruction_rms": 0.0009872300218907086, + "raw_apical_rms": 0.19471755639280275, + "teaching_rms": 0.000987230021848355, + "traffic_rms": 0.19471406160877613 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.039173866956844e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4732077390210166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01610349988045273, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 0.012553250658623874 + }, + { + "epoch": 119, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05110068619251251, + "mean_feedback_forward_cosine": 0.9994942654262889, + "mean_feedback_forward_relative_error": 0.030648558171973986, + "min_feedback_forward_cosine": 0.9986991882324219 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009489401012263225, + "instruction_rms": 0.0009489401012385424, + "raw_apical_rms": 0.1947272120355095, + "teaching_rms": 0.0009489401012263225, + "traffic_rms": 0.19472380274891124 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1731598448250224e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478854731469443e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016094396252802807, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 0.011629069031774998 + }, + { + "epoch": 120, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051095809787511826, + "mean_feedback_forward_cosine": 0.9994943597099998, + "mean_feedback_forward_relative_error": 0.030645027765157547, + "min_feedback_forward_cosine": 0.9986993670463562 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000997190217897782, + "instruction_rms": 0.0009971902179680184, + "raw_apical_rms": 0.19471980221911275, + "teaching_rms": 0.000997190217897782, + "traffic_rms": 0.19471619939068258 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1758748286183618e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4735707785782416e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01610282538963277, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 0.011803273819833994 + }, + { + "epoch": 121, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05109076201915741, + "mean_feedback_forward_cosine": 0.9994944669983604, + "mean_feedback_forward_relative_error": 0.03064156755466353, + "min_feedback_forward_cosine": 0.9986998438835144 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009237097236852764, + "instruction_rms": 0.0009237097238129419, + "raw_apical_rms": 0.19473176158424785, + "teaching_rms": 0.0009237097236852764, + "traffic_rms": 0.19472839854488092 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0096327685005235e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4746056633719105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01598576762070078, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 0.01091268304526806 + }, + { + "epoch": 122, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05108627676963806, + "mean_feedback_forward_cosine": 0.9994945667006753, + "mean_feedback_forward_relative_error": 0.030638265948403964, + "min_feedback_forward_cosine": 0.9986999034881592 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009012204086428553, + "instruction_rms": 0.0009012204085983514, + "raw_apical_rms": 0.19470813452552477, + "teaching_rms": 0.0009012204086428553, + "traffic_rms": 0.1947049196934674 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0666927607871912e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.470975441650929e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01624624478541864, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 0.01006043321505189 + }, + { + "epoch": 123, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05108192190527916, + "mean_feedback_forward_cosine": 0.9994946804913607, + "mean_feedback_forward_relative_error": 0.03063515511087396, + "min_feedback_forward_cosine": 0.9987002611160278 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008771126093473089, + "instruction_rms": 0.0008771126094273887, + "raw_apical_rms": 0.1947415313502428, + "teaching_rms": 0.0008771126093473089, + "traffic_rms": 0.19473850048860603 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0076405487779994e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4766508858289272e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01631890479470542, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 0.009727895104885101 + }, + { + "epoch": 124, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05107719823718071, + "mean_feedback_forward_cosine": 0.9994947661053051, + "mean_feedback_forward_relative_error": 0.030631877414204856, + "min_feedback_forward_cosine": 0.9987005591392517 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008872967030228281, + "instruction_rms": 0.0008872967029505432, + "raw_apical_rms": 0.1946942411192637, + "teaching_rms": 0.0008872967030228281, + "traffic_rms": 0.19469110107709262 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.668006129004425e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4729694016955726e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015812194573190726, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 0.009744029443040491 + }, + { + "epoch": 125, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051072537899017334, + "mean_feedback_forward_cosine": 0.9994948679750616, + "mean_feedback_forward_relative_error": 0.030628488263623282, + "min_feedback_forward_cosine": 0.998700737953186 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008882672810285132, + "instruction_rms": 0.0008882672808986647, + "raw_apical_rms": 0.1947545448404164, + "teaching_rms": 0.0008882672810285132, + "traffic_rms": 0.19475136221868986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0344167833409301e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4720425620664744e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016216655209315284, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 0.009705293390452861 + }, + { + "epoch": 126, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05106782540678978, + "mean_feedback_forward_cosine": 0.9994949633424932, + "mean_feedback_forward_relative_error": 0.030624984653497285, + "min_feedback_forward_cosine": 0.9987010359764099 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008973059666288583, + "instruction_rms": 0.0008973059668333311, + "raw_apical_rms": 0.19479052703081662, + "teaching_rms": 0.0008973059666288583, + "traffic_rms": 0.1947872575090526 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.788552546297069e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4712439766828123e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01619252504669784, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 0.009590714794099332 + }, + { + "epoch": 127, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05106387659907341, + "mean_feedback_forward_cosine": 0.9994950489564376, + "mean_feedback_forward_relative_error": 0.0306220787323334, + "min_feedback_forward_cosine": 0.9987012147903442 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000799109650811655, + "instruction_rms": 0.000799109650702349, + "raw_apical_rms": 0.19477784797611788, + "teaching_rms": 0.000799109650811655, + "traffic_rms": 0.1947751794111017 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.471198580968121e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4733386781525607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016203621218705614, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 0.008340846708118915 + }, + { + "epoch": 128, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05105983838438988, + "mean_feedback_forward_cosine": 0.9994951356541026, + "mean_feedback_forward_relative_error": 0.030619269313121384, + "min_feedback_forward_cosine": 0.9987014532089233 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007730232060228982, + "instruction_rms": 0.0007730232059174983, + "raw_apical_rms": 0.19470011665498896, + "teaching_rms": 0.0007730232060228982, + "traffic_rms": 0.19469762902669124 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1254807219529539e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47790704744727e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015946321393286732, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 0.007661091483235359 + }, + { + "epoch": 129, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051055680960416794, + "mean_feedback_forward_cosine": 0.9994952342726967, + "mean_feedback_forward_relative_error": 0.030616274594583295, + "min_feedback_forward_cosine": 0.9987016916275024 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008044859086443686, + "instruction_rms": 0.0008044859086718653, + "raw_apical_rms": 0.19472318417701792, + "teaching_rms": 0.0008044859086443686, + "traffic_rms": 0.194720541363884 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0069637568221879e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4743581455576497e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01660637242944846, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 0.007918635245561599 + }, + { + "epoch": 130, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05105059593915939, + "mean_feedback_forward_cosine": 0.9994953545657072, + "mean_feedback_forward_relative_error": 0.030612705842676488, + "min_feedback_forward_cosine": 0.9987019300460815 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008934586464189411, + "instruction_rms": 0.0008934586465131052, + "raw_apical_rms": 0.19466745347483663, + "teaching_rms": 0.0008934586464189411, + "traffic_rms": 0.19466418675172056 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.031319669664299e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4707779578040437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01587644993935074, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 0.008843204127773643 + }, + { + "epoch": 131, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05104566738009453, + "mean_feedback_forward_cosine": 0.9994954336773265, + "mean_feedback_forward_relative_error": 0.030609300317750734, + "min_feedback_forward_cosine": 0.9987021088600159 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008166452917566173, + "instruction_rms": 0.0008166452915282878, + "raw_apical_rms": 0.19468633030183363, + "teaching_rms": 0.0008166452917566173, + "traffic_rms": 0.19468348015804474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.3837717460774002e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477716439648477e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01594794643340169, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 0.007759806679040193 + }, + { + "epoch": 132, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051041483879089355, + "mean_feedback_forward_cosine": 0.9994955257935958, + "mean_feedback_forward_relative_error": 0.030606337746774608, + "min_feedback_forward_cosine": 0.9987024068832397 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007524957805809672, + "instruction_rms": 0.0007524957804887153, + "raw_apical_rms": 0.19467457389824397, + "teaching_rms": 0.0007524957805809672, + "traffic_rms": 0.19467202673351666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.15865411487448e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4832584456779786e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01598507454040379, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 0.007225984745770693 + }, + { + "epoch": 133, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051037050783634186, + "mean_feedback_forward_cosine": 0.999495617909865, + "mean_feedback_forward_relative_error": 0.030603039467876608, + "min_feedback_forward_cosine": 0.9987025260925293 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008049028269084094, + "instruction_rms": 0.0008049028270110402, + "raw_apical_rms": 0.1946666405132097, + "teaching_rms": 0.0008049028269084094, + "traffic_rms": 0.19466374059327013 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.287553126569037e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4778675713841723e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015752381001582577, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 0.007561025608740747 + }, + { + "epoch": 134, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05103234201669693, + "mean_feedback_forward_cosine": 0.9994957241145047, + "mean_feedback_forward_relative_error": 0.030599776884032922, + "min_feedback_forward_cosine": 0.998702883720398 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008093043397578135, + "instruction_rms": 0.0008093043398846242, + "raw_apical_rms": 0.19468612894130638, + "teaching_rms": 0.0008093043397578135, + "traffic_rms": 0.19468338433962268 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.477744291378258e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4809717561062823e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015679943666706803, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 0.007401762579381466 + }, + { + "epoch": 135, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05102798342704773, + "mean_feedback_forward_cosine": 0.9994958140633323, + "mean_feedback_forward_relative_error": 0.03059667105871168, + "min_feedback_forward_cosine": 0.9987031817436218 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007527554726908922, + "instruction_rms": 0.0007527554726412799, + "raw_apical_rms": 0.1946748341250857, + "teaching_rms": 0.0007527554726908922, + "traffic_rms": 0.19467234122860363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1290640600236657e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4813215147719377e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015786594480590912, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 0.0066503033824265 + }, + { + "epoch": 136, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051023226231336594, + "mean_feedback_forward_cosine": 0.9994959083470432, + "mean_feedback_forward_relative_error": 0.030593581472269513, + "min_feedback_forward_cosine": 0.9987033605575562 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007620617025212957, + "instruction_rms": 0.0007620617023781779, + "raw_apical_rms": 0.19463765323972385, + "teaching_rms": 0.0007620617025212957, + "traffic_rms": 0.19463495919681986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1380932818383371e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4729208768067923e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015877088238421877, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 0.006470339124202728 + }, + { + "epoch": 137, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051018524914979935, + "mean_feedback_forward_cosine": 0.9994959982958707, + "mean_feedback_forward_relative_error": 0.030590490759773686, + "min_feedback_forward_cosine": 0.9987035393714905 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007537594848222147, + "instruction_rms": 0.0007537594847082503, + "raw_apical_rms": 0.19455945206039038, + "teaching_rms": 0.0007537594848222147, + "traffic_rms": 0.19455700126812453 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1792537435439954e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4726872780437115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016308113849850536, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 0.006562902651429176 + }, + { + "epoch": 138, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05101362615823746, + "mean_feedback_forward_cosine": 0.9994961164214394, + "mean_feedback_forward_relative_error": 0.030587172305042095, + "min_feedback_forward_cosine": 0.9987037777900696 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007293125547284646, + "instruction_rms": 0.0007293125549048965, + "raw_apical_rms": 0.19452935467062965, + "teaching_rms": 0.0007293125547284646, + "traffic_rms": 0.19452684259513653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1010809686240464e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.473961506662821e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016403604987396698, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 0.006227076387871057 + }, + { + "epoch": 139, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05100905895233154, + "mean_feedback_forward_cosine": 0.9994961879470131, + "mean_feedback_forward_relative_error": 0.03058408113535155, + "min_feedback_forward_cosine": 0.9987040162086487 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007464280255729528, + "instruction_rms": 0.0007464280256642835, + "raw_apical_rms": 0.19451711730595664, + "teaching_rms": 0.0007464280255729528, + "traffic_rms": 0.19451470775557805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1225411356008408e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4766619252699227e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015955436771967726, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 0.0063647708220779895 + }, + { + "epoch": 140, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05100460350513458, + "mean_feedback_forward_cosine": 0.9994962952353738, + "mean_feedback_forward_relative_error": 0.030581047021868556, + "min_feedback_forward_cosine": 0.9987043142318726 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007101812682885326, + "instruction_rms": 0.0007101812684891331, + "raw_apical_rms": 0.19458140025569287, + "teaching_rms": 0.0007101812682885326, + "traffic_rms": 0.19457910455763497 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.92375817077118e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4711722491063216e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016494309919204458, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 0.005904565336853265 + }, + { + "epoch": 141, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.051000043749809265, + "mean_feedback_forward_cosine": 0.9994963732632723, + "mean_feedback_forward_relative_error": 0.0305780806240033, + "min_feedback_forward_cosine": 0.9987046122550964 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006831255921249603, + "instruction_rms": 0.0006831255919468898, + "raw_apical_rms": 0.19460190759229162, + "teaching_rms": 0.0006831255921249603, + "traffic_rms": 0.19459958639232044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.3314866009750403e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4762430129526437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016140351180951536, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 0.005442295888029039 + }, + { + "epoch": 142, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509953573346138, + "mean_feedback_forward_cosine": 0.9994964718818664, + "mean_feedback_forward_relative_error": 0.030574862032451414, + "min_feedback_forward_cosine": 0.9987048506736755 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007050308548416913, + "instruction_rms": 0.0007050308547992177, + "raw_apical_rms": 0.19455690181225735, + "teaching_rms": 0.0007050308548416913, + "traffic_rms": 0.19455456402771004 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.2237669011483376e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477201895984719e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016045030915323066, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 0.005739662863835693 + }, + { + "epoch": 143, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05099065229296684, + "mean_feedback_forward_cosine": 0.9994965770027854, + "mean_feedback_forward_relative_error": 0.030571749281476844, + "min_feedback_forward_cosine": 0.9987050294876099 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007072141701633808, + "instruction_rms": 0.0007072141700930291, + "raw_apical_rms": 0.1945340038335725, + "teaching_rms": 0.0007072141701633808, + "traffic_rms": 0.19453173969747178 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0454817100935543e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747661709867035e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016765352910161565, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 0.005758936830982566 + }, + { + "epoch": 144, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050986435264348984, + "mean_feedback_forward_cosine": 0.9994966767051003, + "mean_feedback_forward_relative_error": 0.030568887809799477, + "min_feedback_forward_cosine": 0.9987052083015442 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006303399603426065, + "instruction_rms": 0.0006303399602474879, + "raw_apical_rms": 0.1945006644101127, + "teaching_rms": 0.0006303399603426065, + "traffic_rms": 0.19449862340074567 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1798984189681505e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.487652042605667e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016705521447749342, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 0.00522493914835155 + }, + { + "epoch": 145, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05098193883895874, + "mean_feedback_forward_cosine": 0.9994967644864863, + "mean_feedback_forward_relative_error": 0.030565938201140275, + "min_feedback_forward_cosine": 0.9987055659294128 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006443283719083028, + "instruction_rms": 0.000644328371815122, + "raw_apical_rms": 0.19453178912060828, + "teaching_rms": 0.0006443283719083028, + "traffic_rms": 0.19452976811962555 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0842033759672631e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4788334608011483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016510232063139882, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 0.0050576650580856945 + }, + { + "epoch": 146, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05097801983356476, + "mean_feedback_forward_cosine": 0.9994968370957809, + "mean_feedback_forward_relative_error": 0.03056327812373638, + "min_feedback_forward_cosine": 0.9987057447433472 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006328432623904743, + "instruction_rms": 0.0006328432625319321, + "raw_apical_rms": 0.19450315825525077, + "teaching_rms": 0.0006328432623904743, + "traffic_rms": 0.19450128517702456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0351779167194763e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4748429126422108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016198603061680805, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 0.004863605909571051 + }, + { + "epoch": 147, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509709008038044, + "mean_feedback_forward_cosine": 0.9994969725608825, + "mean_feedback_forward_relative_error": 0.030558950640261172, + "min_feedback_forward_cosine": 0.9987062215805054 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007657752953414205, + "instruction_rms": 0.0007657752952447241, + "raw_apical_rms": 0.1944579500773053, + "teaching_rms": 0.0007657752953414205, + "traffic_rms": 0.19445490136802276 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.934038303072157e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47744352066399e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01633406655955941, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 0.0058586148860678075 + }, + { + "epoch": 148, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050965435802936554, + "mean_feedback_forward_cosine": 0.99949707551436, + "mean_feedback_forward_relative_error": 0.030555595372888177, + "min_feedback_forward_cosine": 0.9987063407897949 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006806242716517995, + "instruction_rms": 0.0006806242715245588, + "raw_apical_rms": 0.1944762104108118, + "teaching_rms": 0.0006806242716517995, + "traffic_rms": 0.19447388186795375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0340680489662191e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4765317144283138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016403460916383773, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 0.005159672223012894 + }, + { + "epoch": 149, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050959449261426926, + "mean_feedback_forward_cosine": 0.9994972001422535, + "mean_feedback_forward_relative_error": 0.03055158338763497, + "min_feedback_forward_cosine": 0.9987066984176636 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007511380318034196, + "instruction_rms": 0.0007511380318617458, + "raw_apical_rms": 0.19450200794285094, + "teaching_rms": 0.0007511380318034196, + "traffic_rms": 0.1944992214377213 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1870838534377981e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4806454133578882e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016424034288043832, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 0.005929333122372627 + }, + { + "epoch": 150, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953879952430725, + "mean_feedback_forward_cosine": 0.9994973041794517, + "mean_feedback_forward_relative_error": 0.030547982979227197, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006894144551137881, + "instruction_rms": 0.0006894144553867142, + "raw_apical_rms": 0.19449710147211996, + "teaching_rms": 0.0006894144551137881, + "traffic_rms": 0.1944946654659506 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.2209676469865371e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478700341916886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015954151218034073, + "minimum_observations": 80 + }, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 0.005376992893218994 + }, + { + "epoch": 151, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095382779836655, + "mean_feedback_forward_cosine": 0.9994973171841014, + "mean_feedback_forward_relative_error": 0.030547930698164485, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005686062400972456, + "instruction_rms": 0.0005686062399989794, + "raw_apical_rms": 0.19439299394352047, + "teaching_rms": 0.0005686062400972456, + "traffic_rms": 0.1943913884213745 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.829823000018223e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787733538164174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016116238840247814, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 0.00433820597358048 + }, + { + "epoch": 152, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953783094882965, + "mean_feedback_forward_cosine": 0.9994973117654974, + "mean_feedback_forward_relative_error": 0.030547880203547803, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006113100864904673, + "instruction_rms": 0.0006113100863838639, + "raw_apical_rms": 0.19438499040309787, + "teaching_rms": 0.0006113100864904673, + "traffic_rms": 0.19438295769604402 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.033216392443137e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475813768372292e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015892574323676547, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 0.004735164075344801 + }, + { + "epoch": 153, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095374956727028, + "mean_feedback_forward_cosine": 0.9994973215189847, + "mean_feedback_forward_relative_error": 0.030547833374955438, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005567400774530025, + "instruction_rms": 0.0005567400774740134, + "raw_apical_rms": 0.194396292280153, + "teaching_rms": 0.0005567400774530025, + "traffic_rms": 0.19439462033802304 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1584438652789686e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4847082602856543e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01607170563152324, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 0.004176526947245002 + }, + { + "epoch": 154, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509536936879158, + "mean_feedback_forward_cosine": 0.9994973161003806, + "mean_feedback_forward_relative_error": 0.03054778448898684, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005408043411764557, + "instruction_rms": 0.0005408043410865101, + "raw_apical_rms": 0.194382386062418, + "teaching_rms": 0.0005408043411764557, + "traffic_rms": 0.19438083221380825 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0227934410522721e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.47703642466691e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016154470518518912, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 0.004001928318999708 + }, + { + "epoch": 155, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095367878675461, + "mean_feedback_forward_cosine": 0.9994973236864263, + "mean_feedback_forward_relative_error": 0.030547738964246077, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005497318664140367, + "instruction_rms": 0.0005497318664043876, + "raw_apical_rms": 0.19439082176146305, + "teaching_rms": 0.0005497318664140367, + "traffic_rms": 0.1943892297359832 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.217525493757421e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4775907383121267e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015871477673435254, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 0.003993319243788719 + }, + { + "epoch": 156, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095362290740013, + "mean_feedback_forward_cosine": 0.9994973280213096, + "mean_feedback_forward_relative_error": 0.030547694218429653, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005009332705483822, + "instruction_rms": 0.0005009332704806486, + "raw_apical_rms": 0.19439572247947026, + "teaching_rms": 0.0005009332705483822, + "traffic_rms": 0.19439431617396907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0780829029499728e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4795967123814573e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016036857931488336, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 0.003720300946496427 + }, + { + "epoch": 157, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953593105077744, + "mean_feedback_forward_cosine": 0.9994973074306142, + "mean_feedback_forward_relative_error": 0.030547652461312035, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004848522280757793, + "instruction_rms": 0.0004848522280396402, + "raw_apical_rms": 0.1944004951303264, + "teaching_rms": 0.0004848522280757793, + "traffic_rms": 0.1943992214213437 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0047433818272111e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4783422124034816e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01608902302900533, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 0.003450639179944992 + }, + { + "epoch": 158, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953567028045654, + "mean_feedback_forward_cosine": 0.9994973226027055, + "mean_feedback_forward_relative_error": 0.030547607664696195, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005563165399655503, + "instruction_rms": 0.0005563165400432616, + "raw_apical_rms": 0.19439335945440914, + "teaching_rms": 0.0005563165399655503, + "traffic_rms": 0.1943917170105444 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.727466476761037e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4787665219808057e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016030034801216904, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 0.00411181521622464 + }, + { + "epoch": 159, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095352604985237, + "mean_feedback_forward_cosine": 0.9994973247701472, + "mean_feedback_forward_relative_error": 0.03054756237701936, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005073773868176134, + "instruction_rms": 0.0005073773867357423, + "raw_apical_rms": 0.194394392463242, + "teaching_rms": 0.0005073773868176134, + "traffic_rms": 0.1943928085240972 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.238599622865877e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4772866714622007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016366823214718645, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 0.003885937132909894 + }, + { + "epoch": 160, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095352232456207, + "mean_feedback_forward_cosine": 0.9994973193515431, + "mean_feedback_forward_relative_error": 0.03054751387204636, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005339157787393562, + "instruction_rms": 0.000533915778573629, + "raw_apical_rms": 0.19439561965426855, + "teaching_rms": 0.0005339157787393562, + "traffic_rms": 0.19439402800510144 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.02918202427827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4804602568018953e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015965250410718446, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 0.003786678851880133 + }, + { + "epoch": 161, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953466445207596, + "mean_feedback_forward_cosine": 0.9994973150166598, + "mean_feedback_forward_relative_error": 0.030547470794144, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00050068567068122, + "instruction_rms": 0.0005006856706463102, + "raw_apical_rms": 0.19439290045267146, + "teaching_rms": 0.00050068567068122, + "traffic_rms": 0.19439157342722688 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.520366717197248e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4804034127835337e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01602907642104826, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 0.0036773203277587893 + }, + { + "epoch": 162, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509534515440464, + "mean_feedback_forward_cosine": 0.9994973106817766, + "mean_feedback_forward_relative_error": 0.030547428368167442, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004745926974471224, + "instruction_rms": 0.0004745926972357706, + "raw_apical_rms": 0.19439609640556044, + "teaching_rms": 0.0004745926974471224, + "traffic_rms": 0.19439487625515162 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.2171129526450386e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477603342782502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01598334415238, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 0.0034456004974618556 + }, + { + "epoch": 163, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095341056585312, + "mean_feedback_forward_cosine": 0.9994973226027055, + "mean_feedback_forward_relative_error": 0.0305473870259117, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048816467371380954, + "instruction_rms": 0.00048816467363634877, + "raw_apical_rms": 0.19438592195417753, + "teaching_rms": 0.00048816467371380954, + "traffic_rms": 0.19438455043582473 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.2595265275194834e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4756499789632602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016110254705855153, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 0.003555494801374152 + }, + { + "epoch": 164, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095338821411133, + "mean_feedback_forward_cosine": 0.9994973421096802, + "mean_feedback_forward_relative_error": 0.030547345615923406, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048881425094263, + "instruction_rms": 0.0004888142507946108, + "raw_apical_rms": 0.19440074789957884, + "teaching_rms": 0.00048881425094263, + "traffic_rms": 0.1943993561785373 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.424202573882212e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476784956707275e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016322284059361376, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 0.0034804721284285188 + }, + { + "epoch": 165, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095336213707924, + "mean_feedback_forward_cosine": 0.9994973247701472, + "mean_feedback_forward_relative_error": 0.03054729922759262, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005073791450165861, + "instruction_rms": 0.0005073791449336735, + "raw_apical_rms": 0.1943987908994404, + "teaching_rms": 0.0005073791450165861, + "traffic_rms": 0.19439727024881642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1713773773180947e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4726563554220765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016047732261041944, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 0.00379402401663363 + }, + { + "epoch": 166, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095334351062775, + "mean_feedback_forward_cosine": 0.9994973312724721, + "mean_feedback_forward_relative_error": 0.0305472592653876, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004853718552617885, + "instruction_rms": 0.0004853718551139234, + "raw_apical_rms": 0.19437975051231032, + "teaching_rms": 0.0004853718552617885, + "traffic_rms": 0.19437829926721123 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0913235115594944e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4797485882619964e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01628890612256642, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 0.003485625109933317 + }, + { + "epoch": 167, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095333606004715, + "mean_feedback_forward_cosine": 0.9994973280213096, + "mean_feedback_forward_relative_error": 0.030547220082106915, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00045500787850319283, + "instruction_rms": 0.0004550078783958457, + "raw_apical_rms": 0.1943812531045656, + "teaching_rms": 0.00045500787850319283, + "traffic_rms": 0.19438014622605992 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0570116870667334e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479518113205874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01597608041126603, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 0.00331782726701349 + }, + { + "epoch": 168, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095330998301506, + "mean_feedback_forward_cosine": 0.999497343193401, + "mean_feedback_forward_relative_error": 0.030547178807583723, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048387401543302044, + "instruction_rms": 0.00048387401533549575, + "raw_apical_rms": 0.19438242849087237, + "teaching_rms": 0.00048387401543302044, + "traffic_rms": 0.19438102558329967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1513434827747915e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4778659509214384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01625966350811703, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 0.003435389999523759 + }, + { + "epoch": 169, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095328018069267, + "mean_feedback_forward_cosine": 0.9994973334399136, + "mean_feedback_forward_relative_error": 0.030547138879244977, + "min_feedback_forward_cosine": 0.998707115650177 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00045383562428293604, + "instruction_rms": 0.0004538356241080158, + "raw_apical_rms": 0.19437763605471434, + "teaching_rms": 0.00045383562428293604, + "traffic_rms": 0.19437647207216072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.741020789988397e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.492504585073857e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016410938776079135, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 0.003363378802500665 + }, + { + "epoch": 170, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953250378370285, + "mean_feedback_forward_cosine": 0.9994973410259593, + "mean_feedback_forward_relative_error": 0.030547099475833502, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004700695696519062, + "instruction_rms": 0.00047006956969305156, + "raw_apical_rms": 0.1943854966722372, + "teaching_rms": 0.0004700695696519062, + "traffic_rms": 0.1943841116045905 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 8.84195472394822e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4794987583875273e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01612671598869285, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 0.0035276851420849562 + }, + { + "epoch": 171, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095319449901581, + "mean_feedback_forward_cosine": 0.9994973280213096, + "mean_feedback_forward_relative_error": 0.03054705672812733, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005238493541702635, + "instruction_rms": 0.0005238493541568089, + "raw_apical_rms": 0.19438317107992117, + "teaching_rms": 0.0005238493541702635, + "traffic_rms": 0.1943816574361072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.62414503646869e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4797149147984774e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01622757517938349, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 0.003712430686634034 + }, + { + "epoch": 172, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095316097140312, + "mean_feedback_forward_cosine": 0.9994973236864263, + "mean_feedback_forward_relative_error": 0.030547016698189756, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004306105991006768, + "instruction_rms": 0.00043061059890058207, + "raw_apical_rms": 0.19437450160401942, + "teaching_rms": 0.0004306105991006768, + "traffic_rms": 0.1943734613213493 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.126703566001197e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4747161691785686e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0162593619881028, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 0.003162156808041036 + }, + { + "epoch": 173, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095311999320984, + "mean_feedback_forward_cosine": 0.999497348612005, + "mean_feedback_forward_relative_error": 0.030546979877081784, + "min_feedback_forward_cosine": 0.9987071752548218 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00043584872559382067, + "instruction_rms": 0.00043584872538832034, + "raw_apical_rms": 0.19437154518328, + "teaching_rms": 0.00043584872559382067, + "traffic_rms": 0.19437042721757275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0339353195831791e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4759153650571056e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016099492258668375, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 0.0030199620487168432 + }, + { + "epoch": 174, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953079015016556, + "mean_feedback_forward_cosine": 0.9994973421096802, + "mean_feedback_forward_relative_error": 0.030546936807645992, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004956851174182257, + "instruction_rms": 0.000495685117368354, + "raw_apical_rms": 0.1943551699309584, + "teaching_rms": 0.0004956851174182257, + "traffic_rms": 0.1943537634709133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0297489438926277e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4796686363598732e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016417340370545493, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 0.0036418039279803634 + }, + { + "epoch": 175, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095306411385536, + "mean_feedback_forward_cosine": 0.9994973247701472, + "mean_feedback_forward_relative_error": 0.030546899309212512, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004424548715424471, + "instruction_rms": 0.0004424548714436922, + "raw_apical_rms": 0.19436383535762555, + "teaching_rms": 0.0004424548715424471, + "traffic_rms": 0.1943627450677169 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1835465585363636e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476477495752896e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016072867900947398, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 0.0032110765231400727 + }, + { + "epoch": 176, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095304921269417, + "mean_feedback_forward_cosine": 0.9994973410259593, + "mean_feedback_forward_relative_error": 0.030546860007399864, + "min_feedback_forward_cosine": 0.998707115650177 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004462428040037558, + "instruction_rms": 0.000446242803926669, + "raw_apical_rms": 0.19436798697011975, + "teaching_rms": 0.0004462428040037558, + "traffic_rms": 0.19436683281745282 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0269862826817189e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4799409695398532e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01675451001148973, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 0.00320290248055011 + }, + { + "epoch": 177, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050953011959791183, + "mean_feedback_forward_cosine": 0.9994973388585178, + "mean_feedback_forward_relative_error": 0.030546818318014795, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004900531342753635, + "instruction_rms": 0.0004900531341510714, + "raw_apical_rms": 0.19436312684314774, + "teaching_rms": 0.0004900531342753635, + "traffic_rms": 0.19436159888017115 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1903637187060667e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4812604164949143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016286448759430246, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 0.003654678724296391 + }, + { + "epoch": 178, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095299705862999, + "mean_feedback_forward_cosine": 0.9994973366910761, + "mean_feedback_forward_relative_error": 0.03054677728055553, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004728026108650528, + "instruction_rms": 0.00047280261078164643, + "raw_apical_rms": 0.19435434882406438, + "teaching_rms": 0.0004728026108650528, + "traffic_rms": 0.19435298688655375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0715405807104617e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4790701760499414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016098918512028007, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 0.0033802037043496967 + }, + { + "epoch": 179, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952985882759094, + "mean_feedback_forward_cosine": 0.9994973399422385, + "mean_feedback_forward_relative_error": 0.03054673769087954, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048392114840683804, + "instruction_rms": 0.00048392114827026033, + "raw_apical_rms": 0.19436654765317193, + "teaching_rms": 0.00048392114840683804, + "traffic_rms": 0.1943650833695793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.34639672323101e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4773118495548762e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01675780583611153, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 0.003343506654240191 + }, + { + "epoch": 180, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095295235514641, + "mean_feedback_forward_cosine": 0.9994973301887512, + "mean_feedback_forward_relative_error": 0.030546696704219688, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004708342637899423, + "instruction_rms": 0.0004708342637267727, + "raw_apical_rms": 0.19436398675108582, + "teaching_rms": 0.0004708342637899423, + "traffic_rms": 0.1943626605796642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.0398630223562577e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4789775363708993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016044407702251975, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 0.0034452605437114833 + }, + { + "epoch": 181, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095292627811432, + "mean_feedback_forward_cosine": 0.9994973377747969, + "mean_feedback_forward_relative_error": 0.03054665649648417, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00046582112671858224, + "instruction_rms": 0.00046582112673901143, + "raw_apical_rms": 0.19435789297602246, + "teaching_rms": 0.00046582112671858224, + "traffic_rms": 0.19435665198401653 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1168807390049551e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478527786206956e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016086212331537477, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 0.003384518695808947 + }, + { + "epoch": 182, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952889025211334, + "mean_feedback_forward_cosine": 0.9994973388585178, + "mean_feedback_forward_relative_error": 0.03054661682214249, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004176398093521574, + "instruction_rms": 0.0004176398093987656, + "raw_apical_rms": 0.1943475528912763, + "teaching_rms": 0.0004176398093521574, + "traffic_rms": 0.1943464797205718 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.717147975152329e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.481905861012614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01641896035340103, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 0.0029255119502544402 + }, + { + "epoch": 183, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095289275050163, + "mean_feedback_forward_cosine": 0.9994973356073553, + "mean_feedback_forward_relative_error": 0.03054657919671048, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004511709365620468, + "instruction_rms": 0.0004511709363906925, + "raw_apical_rms": 0.1943420628582151, + "teaching_rms": 0.0004511709365620468, + "traffic_rms": 0.19434074745974164 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.485943763820615e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479582279332791e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01628925078677219, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 0.003307819086313248 + }, + { + "epoch": 184, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095286667346954, + "mean_feedback_forward_cosine": 0.9994973399422385, + "mean_feedback_forward_relative_error": 0.030546537769789047, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004801691728023221, + "instruction_rms": 0.0004801691726060337, + "raw_apical_rms": 0.19435702675723576, + "teaching_rms": 0.0004801691728023221, + "traffic_rms": 0.19435573115796803 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0328427890726743e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477822749984265e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016136591310228855, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 0.0034362031857296826 + }, + { + "epoch": 185, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095284804701805, + "mean_feedback_forward_cosine": 0.999497348612005, + "mean_feedback_forward_relative_error": 0.030546497773717752, + "min_feedback_forward_cosine": 0.998707115650177 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004384346144854318, + "instruction_rms": 0.0004384346146126119, + "raw_apical_rms": 0.19436570356305613, + "teaching_rms": 0.0004384346144854318, + "traffic_rms": 0.19436464884917407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.039843718028351e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.475807470948572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016146225969448876, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 0.0030020920367632062 + }, + { + "epoch": 186, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952814519405365, + "mean_feedback_forward_cosine": 0.9994973399422385, + "mean_feedback_forward_relative_error": 0.03054645589806817, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004709537453569762, + "instruction_rms": 0.00047095374512234835, + "raw_apical_rms": 0.1943738181005348, + "teaching_rms": 0.0004709537453569762, + "traffic_rms": 0.1943725468392063 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1395709265116238e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4755689300441088e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015914585059106496, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 0.0033384222985431552 + }, + { + "epoch": 187, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952766090631485, + "mean_feedback_forward_cosine": 0.999497343193401, + "mean_feedback_forward_relative_error": 0.03054641798477281, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042143396133202335, + "instruction_rms": 0.000421433961264249, + "raw_apical_rms": 0.1943535981473494, + "teaching_rms": 0.00042143396133202335, + "traffic_rms": 0.194352551583021 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.185641309575658e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4851122546240537e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0164003706593835, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 0.0030099991491809486 + }, + { + "epoch": 188, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509527288377285, + "mean_feedback_forward_cosine": 0.9994973280213096, + "mean_feedback_forward_relative_error": 0.030546377912502396, + "min_feedback_forward_cosine": 0.9987069964408875 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004498255870451366, + "instruction_rms": 0.00044982558712435423, + "raw_apical_rms": 0.19436100398689624, + "teaching_rms": 0.0004498255870451366, + "traffic_rms": 0.19435978356775327 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 1.1431566093733636e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4749870115001635e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01623984403113715, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 0.003160910323560238 + }, + { + "epoch": 189, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095267668366432, + "mean_feedback_forward_cosine": 0.999497348612005, + "mean_feedback_forward_relative_error": 0.030546337603168053, + "min_feedback_forward_cosine": 0.9987069368362427 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00043281698382862445, + "instruction_rms": 0.0004328169837947406, + "raw_apical_rms": 0.19435054967131637, + "teaching_rms": 0.00043281698382862445, + "traffic_rms": 0.19434943922677045 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1301014524178754e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.477941700325129e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.015939358340129484, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 0.003039131107851863 + }, + { + "epoch": 190, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952646881341934, + "mean_feedback_forward_cosine": 0.9994973410259593, + "mean_feedback_forward_relative_error": 0.03054629715836861, + "min_feedback_forward_cosine": 0.9987071752548218 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047730680337186164, + "instruction_rms": 0.00047730680327074293, + "raw_apical_rms": 0.19435680673300626, + "teaching_rms": 0.00047730680337186164, + "traffic_rms": 0.19435541089809455 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1021044343806352e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4782506370561978e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016161737042129247, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 0.003400928650479764 + }, + { + "epoch": 191, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095260217785835, + "mean_feedback_forward_cosine": 0.9994973410259593, + "mean_feedback_forward_relative_error": 0.03054626076058908, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000414267698323551, + "instruction_rms": 0.0004142676982094293, + "raw_apical_rms": 0.1943505591412353, + "teaching_rms": 0.000414267698323551, + "traffic_rms": 0.1943494911517414 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.310453208399849e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.478457445800678e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01642727585049604, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 0.0028851407808810473 + }, + { + "epoch": 192, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952598452568054, + "mean_feedback_forward_cosine": 0.9994973366910761, + "mean_feedback_forward_relative_error": 0.030546225268732418, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00037902483245729206, + "instruction_rms": 0.00037902483239095884, + "raw_apical_rms": 0.19435288559463887, + "teaching_rms": 0.00037902483245729206, + "traffic_rms": 0.19435195657109602 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.572438841587427e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.479850125221969e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01639148621201922, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 0.002640341882184148 + }, + { + "epoch": 193, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952568650245667, + "mean_feedback_forward_cosine": 0.9994973356073553, + "mean_feedback_forward_relative_error": 0.030546185264194555, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004216151707999904, + "instruction_rms": 0.0004216151708215546, + "raw_apical_rms": 0.19435788852336688, + "teaching_rms": 0.0004216151707999904, + "traffic_rms": 0.19435682025595988 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0888079771120829e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4823447781166875e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01639284711967306, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 0.0029772627403959633 + }, + { + "epoch": 194, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952568650245667, + "mean_feedback_forward_cosine": 0.9994973572817716, + "mean_feedback_forward_relative_error": 0.030546146690506825, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000449911380051806, + "instruction_rms": 0.00044991137992028555, + "raw_apical_rms": 0.194362549322696, + "teaching_rms": 0.000449911380051806, + "traffic_rms": 0.1943612927543597 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.1048105363897776e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4752142373748065e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016090296627069487, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 0.003208819755911827 + }, + { + "epoch": 195, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095256119966507, + "mean_feedback_forward_cosine": 0.9994973475282842, + "mean_feedback_forward_relative_error": 0.030546104933389207, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004796360836001401, + "instruction_rms": 0.0004796360834233242, + "raw_apical_rms": 0.19435562752048668, + "teaching_rms": 0.0004796360836001401, + "traffic_rms": 0.19435433381614856 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0522689564140819e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4766122584053554e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016025229980513138, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 0.0032980534649826587 + }, + { + "epoch": 196, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952546298503876, + "mean_feedback_forward_cosine": 0.9994973594492133, + "mean_feedback_forward_relative_error": 0.030546065572310578, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004891858782414675, + "instruction_rms": 0.0004891858780893592, + "raw_apical_rms": 0.19434581537635678, + "teaching_rms": 0.0004891858782414675, + "traffic_rms": 0.19434429506055861 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.323255056870039e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.482799060465587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016181315484985293, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 0.003458643803074956 + }, + { + "epoch": 197, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095251277089119, + "mean_feedback_forward_cosine": 0.9994973518631675, + "mean_feedback_forward_relative_error": 0.030546022198078306, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004495598828716018, + "instruction_rms": 0.00044955988282135366, + "raw_apical_rms": 0.1943482406020158, + "teaching_rms": 0.0004495598828716018, + "traffic_rms": 0.194347062937797 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 9.211665030761651e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4755910373199144e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016095444041352956, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 0.0031894803755916656 + }, + { + "epoch": 198, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0509524829685688, + "mean_feedback_forward_cosine": 0.9994973572817716, + "mean_feedback_forward_relative_error": 0.030545983184129, + "min_feedback_forward_cosine": 0.998707115650177 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00041692178807095274, + "instruction_rms": 0.0004169217879437, + "raw_apical_rms": 0.19435032243613154, + "teaching_rms": 0.00041692178807095274, + "traffic_rms": 0.19434930901499262 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546021699905396, + "maximum_absolute_post_projection_soma_slope": 9.508907794497645e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4809309560621667e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016399476702367903, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 0.0028570882023498416 + }, + { + "epoch": 199, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050952453166246414, + "mean_feedback_forward_cosine": 0.9994973464445635, + "mean_feedback_forward_relative_error": 0.030545949521051212, + "min_feedback_forward_cosine": 0.9987070560455322 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00039462591976481765, + "instruction_rms": 0.00039462591972123115, + "raw_apical_rms": 0.19434591319537498, + "teaching_rms": 0.00039462591976481765, + "traffic_rms": 0.1943449435152257 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.2313265074226365e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.4765461320585462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01593207283687475, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 0.002666809372249991 + }, + { + "epoch": 200, + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05095241591334343, + "mean_feedback_forward_cosine": 0.9994973464445635, + "mean_feedback_forward_relative_error": 0.03054591358893297, + "min_feedback_forward_cosine": 0.9987071752548218 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000409122590855529, + "instruction_rms": 0.00040912259070394294, + "raw_apical_rms": 0.1943344283276579, + "teaching_rms": 0.000409122590855529, + "traffic_rms": 0.19433331417969973 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.36546018719673157, + "maximum_absolute_post_projection_soma_slope": 1.0714307308035131e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.476970796404484e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.016329289586733298, + "minimum_observations": 80 + }, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 0.002732656699893996 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.9274, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 0.37942074279785154, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 3324800000, + "peak_memory_reserved_bytes": 5553258496, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 1.2290752238186542e-07, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 1.2290752238186542e-07, + "min_residual_soma_slope": -1.1848618441945291e-07, + "mse": 1.5285795364643764e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.081359340818409e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.081359340818409e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 2.026231050491333, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.07117486000061035, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 26377.442036390305, + "train_wall_s": 26370.92627930641 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.05091153830289841, + 0.03655247017741203, + 0.023153452202677727, + 0.021897587925195694, + 0.013579683378338814, + 0.015431691892445087, + 0.00934370793402195, + 0.011485620401799679, + 0.007140439935028553, + 0.009418174624443054, + 0.005642977077513933, + 0.007607867009937763, + 0.004767856560647488, + 0.006615776102989912, + 0.003951891791075468, + 0.005578398704528809, + 0.003516203025355935, + 0.0051722414791584015, + 0.003137255320325494, + 0.007752435747534037, + 0.004813292995095253, + 0.006730418652296066, + 0.004173689056187868, + 0.005761435721069574, + 0.00367356208153069, + 0.005273146089166403, + 0.003168795257806778, + 0.004652910400182009, + 0.0027767426799982786, + 0.004038074519485235, + 0.0025574308820068836, + 0.0035546140279620886, + 0.002322361571714282, + 0.0034247329458594322, + 0.0021392563357949257, + 0.003221050603315234, + 0.001980266999453306, + 0.004863648675382137, + 0.0033028495963662863, + 0.004508663900196552, + 0.0030757742933928967, + 0.004294773563742638, + 0.0028490161057561636, + 0.004082180093973875, + 0.002660284051671624, + 0.003812672570347786, + 0.002448310377076268, + 0.0034677244257181883, + 0.00229784962721169, + 0.0034069865942001343, + 0.0021700409706681967, + 0.003208857960999012, + 0.002054570009931922, + 0.0029834227170795202, + 0.0020505560096353292 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.2741865813732147, + 0.19733789563179016, + 0.0829969048500061, + 0.11605869978666306, + 0.03608892858028412, + 0.0833565816283226, + 0.020584456622600555, + 0.062496367841959, + 0.014349586330354214, + 0.04949149489402771, + 0.010205832310020924, + 0.039496153593063354, + 0.007967822253704071, + 0.035705070942640305, + 0.006037096492946148, + 0.028738010674715042, + 0.005009029991924763, + 0.027241500094532967, + 0.004237405955791473, + 0.04153915494680405, + 0.008449469693005085, + 0.03575325012207031, + 0.006828334182500839, + 0.030518390238285065, + 0.00560279143974185, + 0.028213337063789368, + 0.004507121630012989, + 0.02500816248357296, + 0.0037638831418007612, + 0.021143373101949692, + 0.0032912753522396088, + 0.01856883056461811, + 0.0028557528275996447, + 0.018206365406513214, + 0.0024693470913916826, + 0.016764724627137184, + 0.00216887379065156, + 0.026165900751948357, + 0.004856692161411047, + 0.02404988557100296, + 0.004315375350415707, + 0.022511620074510574, + 0.003794856136664748, + 0.021572861820459366, + 0.003355208318680525, + 0.019874075427651405, + 0.002985555212944746, + 0.018577327951788902, + 0.002686375053599477, + 0.01843234896659851, + 0.002411630004644394, + 0.017050232738256454, + 0.0022508215624839067, + 0.01570250280201435, + 0.0021455418318510056 + ], + "unscaled_traffic_rms": [ + 0.7427284121513367, + 0.7409113049507141, + 1.1158705949783325, + 0.7547073364257812, + 1.5051356554031372, + 0.7405146360397339, + 1.8156821727752686, + 0.7351224422454834, + 1.9904239177703857, + 0.7611954212188721, + 2.211667537689209, + 0.7704919576644897, + 2.3935556411743164, + 0.7411581873893738, + 2.61840558052063, + 0.7764488458633423, + 2.807891368865967, + 0.7594649791717529, + 2.96148681640625, + 0.7465183734893799, + 2.278625011444092, + 0.752985417842865, + 2.4449236392974854, + 0.7551428079605103, + 2.6226656436920166, + 0.7476103901863098, + 2.812256336212158, + 0.7442227005958557, + 2.9509339332580566, + 0.7639414072036743, + 3.108133554458618, + 0.7657163143157959, + 3.2528886795043945, + 0.7524254322052002, + 3.465298652648926, + 0.7685304880142212, + 3.6521570682525635, + 0.7435094714164734, + 2.7202460765838623, + 0.7498853206634521, + 2.8509912490844727, + 0.7631211876869202, + 3.0030293464660645, + 0.7569102644920349, + 3.171527624130249, + 0.7673659920692444, + 3.280207872390747, + 0.7466573119163513, + 3.42148756980896, + 0.7393494248390198, + 3.5992932319641113, + 0.7528009414672852, + 3.651235580444336, + 0.7599865794181824, + 3.8229150772094727 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1250441282772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1250433280000000, + "local_weight_correlation_macs": 1254856960000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1254856960000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 127796290519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 10000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000064, + "total_forward_equivalent_examples": 10000064, + "total_macs_estimate": 5010596513857536 + } +} diff --git a/results/oral_a_dynamic_scaling_v2_gate.json b/results/oral_a_dynamic_scaling_v2_gate.json new file mode 100644 index 0000000..f1040ba --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2_gate.json @@ -0,0 +1,434 @@ +{ + "analyzer_sha256": "ee911f02f5ad75a5b2c233878b6b617b7652cfc3615335f6c9cffc33aed931c3", + "bci_v2_gate_sha256": "3b161c32727a2a7c4b50ca696dafff20a023d23e8cd008a0b8dddd662516b084", + "bci_v2_protocol": "oral_b_v2_calibrated_recovery_confirmation_v1", + "checks": { + "all_60_records_and_audited_values_finite": true, + "all_mechanism_query_cost_memory_invariants": true, + "at_least_four_dynamic_seeds_improve_with_depth": true, + "bp_mean_accuracy_at_least_0p90_each_depth": true, + "d56_dynamic_dfa_advantage_lower_bound_at_least_1pt": true, + "d56_dynamic_mean_advantage_over_dfa_at_least_2pt": true, + "d56_mean_alignment_retains_90pct_of_d20": true, + "dynamic_bp_deficit_upper_bound_at_most_3pt_each_depth": true, + "dynamic_depth_gain_lower_bound_nonnegative": true, + "dynamic_depth_gain_within_1pt_of_bp_gain": true, + "dynamic_kp_deficit_upper_bound_at_most_2p5pt_each_depth": true, + "dynamic_mean_alignment_at_least_0p85_each_depth": true, + "dynamic_mean_d20_to_d56_gain_at_least_0p5pt": true, + "dynamic_mean_within_1p5pt_kp_each_depth": true, + "dynamic_mean_within_2pt_bp_each_depth": true, + "every_d56_dynamic_alignment_at_least_0p80": true, + "every_dynamic_accuracy_at_least_0p87": true + }, + "complete_grid": true, + "conv_run_sha256": "359799cef77a2b33dbbeda43e602f92005685ca16b119c4ccd8c076dc176772f", + "d4_gate_sha256": "636c587e47287338cdca7d9558dc3bfb99a2cec615badb23337025d85b8128ef", + "d4_source_commit": "0008f2cd96c06bea98f43560867d93967c597711", + "legacy_analyzer_sha256": "6cca23d01895e8704f8cb9e65fa6814bce4d193b4d6a14cb9723ac3f911df656", + "legacy_runner_sha256": "f7762dfaa764f6ae7e73dfb4780af3bc7fcaa02f94dee278906b43830367b399", + "metrics": { + "accuracy_by_method_depth_seed": { + "bp": { + "20": [ + 0.9139, + 0.9188, + 0.9142, + 0.9163, + 0.918 + ], + "32": [ + 0.9179, + 0.9266, + 0.9228, + 0.9262, + 0.9216 + ], + "56": [ + 0.9196, + 0.9275, + 0.9288, + 0.9306, + 0.9251 + ] + }, + "clean_kp": { + "20": [ + 0.9158, + 0.9155, + 0.9087, + 0.915, + 0.9144 + ], + "32": [ + 0.9223, + 0.9242, + 0.9209, + 0.9234, + 0.9258 + ], + "56": [ + 0.9287, + 0.9298, + 0.9265, + 0.9273, + 0.9212 + ] + }, + "dfa": { + "20": [ + 0.372, + 0.3374, + 0.2628, + 0.3417, + 0.28 + ], + "32": [ + 0.3521, + 0.3264, + 0.3404, + 0.2863, + 0.329 + ], + "56": [ + 0.3348, + 0.2604, + 0.3382, + 0.3488, + 0.2603 + ] + }, + "dynamic": { + "20": [ + 0.9152, + 0.9158, + 0.9165, + 0.9151, + 0.9166 + ], + "32": [ + 0.9229, + 0.9226, + 0.921, + 0.9217, + 0.9245 + ], + "56": [ + 0.9292, + 0.9285, + 0.9264, + 0.9265, + 0.9274 + ] + } + }, + "bp_d20_to_d56_gain": [ + 0.005699999999999927, + 0.008700000000000041, + 0.014599999999999946, + 0.01429999999999998, + 0.007099999999999995 + ], + "bp_deficit": { + "20": [ + -0.0012999999999999678, + 0.0030000000000000027, + -0.0022999999999999687, + 0.0011999999999999789, + 0.0014000000000000679 + ], + "32": [ + -0.0050000000000000044, + 0.0040000000000000036, + 0.0017999999999999128, + 0.0045000000000000595, + -0.0029000000000000137 + ], + "56": [ + -0.009600000000000053, + -0.0010000000000000009, + 0.0023999999999999577, + 0.0040999999999999925, + -0.0022999999999999687 + ] + }, + "d56_dfa_advantage": [ + 0.5944, + 0.6680999999999999, + 0.5882000000000001, + 0.5777, + 0.6671 + ], + "dynamic_alignment": { + "20": [ + 0.9996865193049113, + 0.9996783634026846, + 0.9997114737828573, + 0.9996656179428101, + 0.9996906618277231 + ], + "32": [ + 0.9995589315891266, + 0.9995903253555298, + 0.9996383130550385, + 0.9996227920055389, + 0.9996523022651672 + ], + "56": [ + 0.9994682272275289, + 0.9994528922769759, + 0.9993400242593553, + 0.9994251562489403, + 0.9994279046853384 + ] + }, + "dynamic_d20_to_d56_gain": [ + 0.014000000000000012, + 0.012700000000000045, + 0.00990000000000002, + 0.011399999999999966, + 0.010800000000000032 + ], + "invariant_failures": [], + "kp_deficit": { + "20": [ + 0.0005999999999999339, + -0.00029999999999996696, + -0.007800000000000029, + -9.999999999998899e-05, + -0.0021999999999999797 + ], + "32": [ + -0.0006000000000000449, + 0.0016000000000000458, + -9.999999999998899e-05, + 0.0017000000000000348, + 0.0012999999999999678 + ], + "56": [ + -0.000500000000000056, + 0.0012999999999999678, + 9.999999999998899e-05, + 0.0008000000000000229, + -0.006199999999999983 + ] + }, + "mean_accuracy": { + "bp": { + "20": 0.91624, + "32": 0.92302, + "56": 0.92632 + }, + "clean_kp": { + "20": 0.91388, + "32": 0.92332, + "56": 0.9267 + }, + "dfa": { + "20": 0.31878, + "32": 0.32684, + "56": 0.3085 + }, + "dynamic": { + "20": 0.91584, + "32": 0.92254, + "56": 0.9276 + } + }, + "mechanism": { + "d20_s10": { + "early_alignment": 0.9996865193049113, + "final_feedback_cosine": 0.9998452883017691, + "instruction_ratio_error": 7.376647070955755e-10, + "last50_feedback_cosine": 0.9998452330890455, + "post_projection_soma_slope": 3.345177290725587e-08, + "post_projection_traffic_ratio": 3.020465978422986e-08 + }, + "d20_s11": { + "early_alignment": 0.9996783634026846, + "final_feedback_cosine": 0.9998476599392138, + "instruction_ratio_error": 9.601693884420115e-10, + "last50_feedback_cosine": 0.9998476189061215, + "post_projection_soma_slope": 4.183369384236357e-08, + "post_projection_traffic_ratio": 3.046251457654459e-08 + }, + "d20_s12": { + "early_alignment": 0.9997114737828573, + "final_feedback_cosine": 0.99984504988319, + "instruction_ratio_error": 7.572193982952058e-10, + "last50_feedback_cosine": 0.9998450080344552, + "post_projection_soma_slope": 6.702554600224175e-08, + "post_projection_traffic_ratio": 3.0341414978753644e-08 + }, + "d20_s13": { + "early_alignment": 0.9996656179428101, + "final_feedback_cosine": 0.9998400305446825, + "instruction_ratio_error": 6.523915851985862e-10, + "last50_feedback_cosine": 0.9998400152357001, + "post_projection_soma_slope": 2.3063201126660715e-08, + "post_projection_traffic_ratio": 3.022787119052029e-08 + }, + "d20_s14": { + "early_alignment": 0.9996906618277231, + "final_feedback_cosine": 0.9998461353151422, + "instruction_ratio_error": 6.969612664775582e-10, + "last50_feedback_cosine": 0.9998460979838121, + "post_projection_soma_slope": 5.1318746585593544e-08, + "post_projection_traffic_ratio": 3.0321173518936684e-08 + }, + "d32_s10": { + "early_alignment": 0.9995589315891266, + "final_feedback_cosine": 0.9997495874281852, + "instruction_ratio_error": 6.764921955948466e-10, + "last50_feedback_cosine": 0.9997495515500345, + "post_projection_soma_slope": 4.883540682953935e-08, + "post_projection_traffic_ratio": 3.032019977820027e-08 + }, + "d32_s11": { + "early_alignment": 0.9995903253555298, + "final_feedback_cosine": 0.9997563823576896, + "instruction_ratio_error": 4.5648018698329906e-10, + "last50_feedback_cosine": 0.9997563595541062, + "post_projection_soma_slope": 6.71051907374931e-08, + "post_projection_traffic_ratio": 3.045556571320131e-08 + }, + "d32_s12": { + "early_alignment": 0.9996383130550385, + "final_feedback_cosine": 0.9997487606540802, + "instruction_ratio_error": 4.706435241530471e-10, + "last50_feedback_cosine": 0.9997487456567826, + "post_projection_soma_slope": 5.9466003676789114e-08, + "post_projection_traffic_ratio": 3.008953587271665e-08 + }, + "d32_s13": { + "early_alignment": 0.9996227920055389, + "final_feedback_cosine": 0.9997596452313084, + "instruction_ratio_error": 6.859344203746787e-10, + "last50_feedback_cosine": 0.9997596325412873, + "post_projection_soma_slope": 4.992172009110618e-08, + "post_projection_traffic_ratio": 3.023138328154589e-08 + }, + "d32_s14": { + "early_alignment": 0.9996523022651672, + "final_feedback_cosine": 0.9997503872840635, + "instruction_ratio_error": 8.423541864033268e-10, + "last50_feedback_cosine": 0.9997503573663773, + "post_projection_soma_slope": 1.3652264385655144e-07, + "post_projection_traffic_ratio": 3.052199008723502e-08 + }, + "d56_s10": { + "early_alignment": 0.9994682272275289, + "final_feedback_cosine": 0.9994443373246626, + "instruction_ratio_error": 1.8433955473184938e-09, + "last50_feedback_cosine": 0.9994443135044792, + "post_projection_soma_slope": 1.2300030505230097e-07, + "post_projection_traffic_ratio": 3.0222815594380756e-08 + }, + "d56_s11": { + "early_alignment": 0.9994528922769759, + "final_feedback_cosine": 0.9995171655308116, + "instruction_ratio_error": 1.0462242183706394e-09, + "last50_feedback_cosine": 0.9995171439647674, + "post_projection_soma_slope": 1.3774025831025938e-07, + "post_projection_traffic_ratio": 3.038032455888949e-08 + }, + "d56_s12": { + "early_alignment": 0.9993400242593553, + "final_feedback_cosine": 0.9995060747320001, + "instruction_ratio_error": 7.430491777427051e-10, + "last50_feedback_cosine": 0.9995060457099568, + "post_projection_soma_slope": 2.3538157734037668e-07, + "post_projection_traffic_ratio": 3.030163009286038e-08 + }, + "d56_s13": { + "early_alignment": 0.9994251562489403, + "final_feedback_cosine": 0.999431696805087, + "instruction_ratio_error": 7.923348643856798e-10, + "last50_feedback_cosine": 0.9994316841689023, + "post_projection_soma_slope": 1.0941191419533425e-07, + "post_projection_traffic_ratio": 3.0173287463056015e-08 + }, + "d56_s14": { + "early_alignment": 0.9994279046853384, + "final_feedback_cosine": 0.9994973464445635, + "instruction_ratio_error": 4.981972612227992e-10, + "last50_feedback_cosine": 0.999497334393588, + "post_projection_soma_slope": 1.477744291378258e-07, + "post_projection_traffic_ratio": 3.031357188939468e-08 + } + } + }, + "new_source_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "old_oral_a_failures_preserved": true, + "protocol": "oral_a_dynamic_innovation_scaling_recovery_v2", + "protocol_sha256": "fb356372d19b09ecdcee639d6521e6922e49437de16e946d16b8bfc623a67b1f", + "review_score_after": 9, + "review_score_before": 8, + "runner_sha256": "72fd8f816d62f11d447c473ecaf08c2ce647e815a7be3b9303e8f34e8d2acd61", + "score_change_rule": "only the complete calibrated-BCI-v2-gated 60-cell standard-ResNet panel establishes oral-A scaling", + "source_sha256": { + "results/kp_dynamic_projection_confirmation/seed10_clean_kp.json": "7c34a0e561446ad6e46575915fb8bb9cddd7b3c24c323bac1e64fea42758eb1f", + "results/kp_dynamic_projection_confirmation/seed10_dynamic.json": "dd74ca4f994e1274bf5d3b6fcc315f73b86fe0ff14c403b0f46868d476f0b5f1", + "results/kp_dynamic_projection_confirmation/seed11_clean_kp.json": "c6449d310c507a03e867046b71d280336c17c94e85ba0335cf2581389f1327ab", + "results/kp_dynamic_projection_confirmation/seed11_dynamic.json": "6a87bccb3db5fab751925d337261a19ddd08604b93f049ad80cc4e6d919fdc75", + "results/kp_dynamic_projection_confirmation/seed12_clean_kp.json": "16341d64a9ec0ad98e635d6c387effab27f4dd2dd63d218f677a8a1716d3dfae", + "results/kp_dynamic_projection_confirmation/seed12_dynamic.json": "337090a537034407a0d54b5c5a27c7502c75770814ad456cadfb9b609977a9a1", + "results/kp_dynamic_projection_confirmation/seed13_clean_kp.json": "e35f0eae6fcd41f2f9731a78eacad572e8b54f79261839faa66b0bd398c23a35", + "results/kp_dynamic_projection_confirmation/seed13_dynamic.json": "047ad3c1a724bdb5b8cdcc88b6c2ee371daa3477db1d7c8f06952aaa6c72822f", + "results/kp_dynamic_projection_confirmation/seed14_clean_kp.json": "500b22ca5cba0dcfd498eb052ac8559e04e750eb10a1945b3a33a52a5f4379eb", + "results/kp_dynamic_projection_confirmation/seed14_dynamic.json": "70221abccb47fe43395936be4963a5f7737d1267f2d9670216d565fcff2ed976", + "results/oral_a_dynamic_scaling_v2/bp_d20_s10.json": "e3d21e970325055bca6a30d9339a84733a9a651bee9b0c13a905e84398b55b1b", + "results/oral_a_dynamic_scaling_v2/bp_d20_s11.json": "9e296e373e3008f9d699a5a67d75fd39fd0db3c5d46ece196ccc4299eff825f8", + "results/oral_a_dynamic_scaling_v2/bp_d20_s12.json": "e17cc41470d4528988720d44704cc1c75bcd9002d1d832cf1aeca20982ff00cf", + "results/oral_a_dynamic_scaling_v2/bp_d20_s13.json": "d978efcda68a316bada66168a0dbf30ac0f2df49ec842e00799a6bff3f19170f", + "results/oral_a_dynamic_scaling_v2/bp_d20_s14.json": "36ccce2890b4c3c9eb1846ef7d3077ac1acab4ab23630f58b690e41d0d218d32", + "results/oral_a_dynamic_scaling_v2/bp_d32_s10.json": "e169abf103873bf7e894f3ab0023aeb716f61591ade34ae6adc8397eb1dd7f2f", + "results/oral_a_dynamic_scaling_v2/bp_d32_s11.json": "0a1d9e8a8f4333660048b8ffe40aca95eaef1a5bdbb5a0585920d100c68d5d18", + "results/oral_a_dynamic_scaling_v2/bp_d32_s12.json": "15cd307b0e1da72a7f830beaf85eb4b7fa9d06653eeb4112c52a3b31d80b692a", + "results/oral_a_dynamic_scaling_v2/bp_d32_s13.json": "fa25dfe4370330cce279b752d01b126331cc4dbaf02d13b703e041e24e59517f", + "results/oral_a_dynamic_scaling_v2/bp_d32_s14.json": "296c6ba300610b59552d41dc99bf0ec1db8013bf8078c6537d4d3600cfac50cc", + "results/oral_a_dynamic_scaling_v2/bp_d56_s10.json": "716656729a89851a0cc47a65998b4f7078b7f4fec8f65459ac24606265cc95a4", + "results/oral_a_dynamic_scaling_v2/bp_d56_s11.json": "a6a5a70e6c3d4a4c151688cee6b1b89fcb67ca7a77b9632af87e020f7cec9a04", + "results/oral_a_dynamic_scaling_v2/bp_d56_s12.json": "ac9176d00f3a7dc0585de5158b97945b36fc773458372d6b012e837e4973dc03", + "results/oral_a_dynamic_scaling_v2/bp_d56_s13.json": "cba10b0c9295d0d8a30e2b34f7fa104c509169984b15a51e3c8022e25927c2d0", + "results/oral_a_dynamic_scaling_v2/bp_d56_s14.json": "23c7221f22e9cccb76bcaa65c394a2b6e77bbc1661c024790e89160b165a7b54", + "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s10.json": "53e629f95e40c0a09bf2e40e5ceb540b3381b7f526a54849b3b746a620fc259a", + "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s11.json": "d21c12da44e6311db5c97ae2538f7b7b349059376edb16ea0e72653fe969f876", + "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s12.json": "56b07513c143a3b148ccf2d9efa999006175ba5ea6c34a542eb00dd6c3470f29", + "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json": "4b33814d71143f8e3add3ae5033a4043940b37226c256c43a36f1c9fb313967f", + "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json": "af797d5586013b0e4db6cb9893c2ba8c0343cfc270221691eb7aff1fb3fb61c6", + "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s10.json": "8f5a9796df26b72cb294a230dce365f56f751110d26aa1b69812ad32230460e8", + "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json": "21bda94033b8a00e0a2613cca189202a09e6a989df45bc728ec3238cf6167045", + "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s12.json": "f294fa6c9be74de1d5da3cb691f79a36ca3c91b3d024f20fe0795cb27203cf0b", + "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s13.json": "8e81eae524565e18661f75ba357e4c2b4413cc7a718d539e7bd0a06dd25027e9", + "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s14.json": "dba34e1ab4442686a1a5a60c7a136ca18fdc823758defb8fcafd0955461eb6d3", + "results/oral_a_dynamic_scaling_v2/dfa_d20_s10.json": "cdc2c819abe6416d34814e12d5aa15965ffc2c10ce7f64411e42a115e882efcb", + "results/oral_a_dynamic_scaling_v2/dfa_d20_s11.json": "cf7ca71ed75e59b1160a0e9220e954fbccc1fd3de3d2fc58584ccf18da613ec3", + "results/oral_a_dynamic_scaling_v2/dfa_d20_s12.json": "add58ff4afd96948e08858ec53ce1ae4e1e10002e84fd9f5387b75428887c13e", + "results/oral_a_dynamic_scaling_v2/dfa_d20_s13.json": "8e1f4fa5b7bdd23b71dfc67025dd81cb3620f42e575b017e240e953838e7b30c", + "results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json": "876eba5327ba132af098d0872d97fad3f4801ff93a1aa5c3b85f1ee34698403c", + "results/oral_a_dynamic_scaling_v2/dfa_d32_s10.json": "af04b6f87409a763d45bfe9253e11bde704db32400779b551917ed36b92035f8", + "results/oral_a_dynamic_scaling_v2/dfa_d32_s11.json": "69087c4efa45cd13144320b547f73a5a0bcc4d3588a6fc223b10a2600a527b7e", + "results/oral_a_dynamic_scaling_v2/dfa_d32_s12.json": "f03356aa2e1761c1059148fc7d6453c1e1f568da960b50dc5610e791a8c0d937", + "results/oral_a_dynamic_scaling_v2/dfa_d32_s13.json": "1958799ce9eccb07515f72b1caa6750263528a4c520d388c44d23fb8e22b5277", + "results/oral_a_dynamic_scaling_v2/dfa_d32_s14.json": "7f3f8e890fd31e43fe46d404b4b7e1f0c3f10b4280cef6040682281080d11cd5", + "results/oral_a_dynamic_scaling_v2/dfa_d56_s10.json": "53bb57e35b7b7eb4b61b7eed92b759f99f3bd3c4fcb228fa6ee631b88a6d0a57", + "results/oral_a_dynamic_scaling_v2/dfa_d56_s11.json": "85caa64b24dfdab964a09318b09c875c044480cfe195de37ec79e5f0a1da37c6", + "results/oral_a_dynamic_scaling_v2/dfa_d56_s12.json": "55099bfe04e25914f25a12f016776a9153e68b0ab47192b2db6abbf501d1da9d", + "results/oral_a_dynamic_scaling_v2/dfa_d56_s13.json": "eb78c3653021cc0c746abaf12c1e4c2c8fdfb8463759bd990d27cd802b7aacd3", + "results/oral_a_dynamic_scaling_v2/dfa_d56_s14.json": "b87a06eb5cba577cbc03c22c149fa6a801291532f1af8a755d7f8321dd2005be", + "results/oral_a_dynamic_scaling_v2/dynamic_d32_s10.json": "f482ea1d7e9d1f88ed2f80aa9a62c80d5e59b5785f10bd45860cdec6068b6529", + "results/oral_a_dynamic_scaling_v2/dynamic_d32_s11.json": "ea6c562f14bd9b1b13c7b77b9179e69d8e22abd7c603109cfb041105871b47dd", + "results/oral_a_dynamic_scaling_v2/dynamic_d32_s12.json": "1a7648b264feb357f240ba3e1267e7998484bc0ebbecd368a5f10b1baa890336", + "results/oral_a_dynamic_scaling_v2/dynamic_d32_s13.json": "e7110e05ed867f71c36e0d6db10b8207a5c9b887dbc726e67b9832d4ff76f52b", + "results/oral_a_dynamic_scaling_v2/dynamic_d32_s14.json": "d5217c25ef43f535f0fde0919a4efc17b4fcfadad4cea0fb14a43dc981c72346", + "results/oral_a_dynamic_scaling_v2/dynamic_d56_s10.json": "9f9691038b9f278947080adf7dd83638d72dddd8f4f50fae6e0a584ae2a2963a", + "results/oral_a_dynamic_scaling_v2/dynamic_d56_s11.json": "293bd719938dc5baeb5c609071f592878add932daa0de5f26d767ba630052af5", + "results/oral_a_dynamic_scaling_v2/dynamic_d56_s12.json": "e0aa30d31a32d9ac79ee152fdbea2a564d9fdc57fef931292e0fdaebc886d63a", + "results/oral_a_dynamic_scaling_v2/dynamic_d56_s13.json": "ee3d12f66ef7862af64d2ab4febdb6defddc3fbbb85002bbf5e41fe38388c3a2", + "results/oral_a_dynamic_scaling_v2/dynamic_d56_s14.json": "ce18a0e78ff2437912ae5bf6a6244664e79e086aa3814fc78878532993f870e8" + }, + "standard_depth_scaling_established": true, + "status": "passed", + "test_evaluations": 60 +} -- cgit v1.2.3